2010 PDF - SigBovik

Transcription

2010 PDF - SigBovik

A Record of The Proceedings
of SIGBOVIK 2010
April 1st, 2010
Carnegie Mellon University
Pittsburgh, Pennsylvania, USA*
http://sigbovik.org/2010
*
now available with 10% more healthcare
I
ach
Association for
Computational Heresy
Advancing Computing as Tomfoolery & Distraction
— or —
Hey, Kids, Get Off My Lawn
DISCLAIMER: LATEX Allergies
In the processing of LATEX into commercial documents, chemicals are added which may produce allergic
reactions in certain people. Mild reactions simply appear as dry, itchy skin or a mild rash. Severe
reactions, however, can be life threatening, often involving respiratory difficulties or even shock. These
symptoms may not appear until several hours after exposure. A physician should always be consulted
even for minor symptoms, since further exposure could result in a more serious allergic reaction.
Anyone can develop an allergy to LATEX, but those with ongoing exposure are at highest risk. This group
of high risk employees includes mathematicians, theoretical computer scientists or anyone else who
regularly uses or is exposed to LATEX. People with other allergies have increased risk, especially if those
allergies are related to parameter passing mechanisms or lack of data abstraction.
Most people think of higher mathematics when they think of LATEX, but LATEX may also be present in
many office and household products, including conference proceedings, software manuals, instruction
booklets, grant proposals, and even campus mail. Individuals with a LATEX allergy should be aware
of any products that may trigger a reaction. Always check your mime type before opening any e-mail
attachments!
The best way to avoid an allergic reaction to LATEX is to use LATEX-free products where possible, wear
specifically designed LATEX-filter safety goggles, or use barrier protection. Also, avoid direct sunlight
if you must read LATEX documents (sunlight enhances poor typeface choices and ligatures), avoid the
printer rooms whenever possible, and make sure coworkers and health care providers know of your
allergy. Always wash your hands after leaving Emacs LATEX mode, and if you already have the allergy,
wear a medical alert bracelet.
- Pfenning, Pfenning, and Pfenning, pro-bono legal council for the Association for Computational Heresy
II
A Message
From The Organizing Committee
We are honored, delighted, excited, and amused to present to you this, the record of the Fourth
Annual Intercalary Workshop about Symposium on Robot Dance Party of Conference in
Celebration of Harry Q. Bovik’s (26)th birthday. And what a wonderful fourth year it’s been!
SIGBOVIK is exactly as strong as ever, with exactly more or less precisely the same number of
papers submitted and accepted as last year.
In the truest echo of the ideals underlying SIGBOVIK, it’s been a landmark year for
collaboration, with vastly more papers by multiple authors, many of them based on multi-person
projects, some of them dangerously close to resembling serious research. At the far end of
high collaboration, this year’s program includes the first ever SIGBOVIK paper with more than
ten authors! Let us hope that the glorious tradition of comradely camera-ready camaraderie
continues without end. Thank you all for making SIGBOVIK the indescribable something that
it is.
As in past years, we have taken the liberty of dividing the proceedings into coherent subjectspecific tracks for the purposes of easy access and clarity of reference. This process always
involves a certain amount of favoritism and unfairness, though, as some papers must appear in
the coveted early tracks of the program while other papers necessarily occupy the inevitable but
unenviable role of filling out the final track. But denizens of Academia, friends of Bovik, fear
not! In order to combat this rampant unjustness, we have chosen *not* to order the tracks this
year: to us, every track is Track 1! The Track 1s appear in a random order unique to each copy
of the proceedings, as guaranteed by the following true random number generator:
fun rand () = 100 (* chosen by random.org, using noises from the
atmosphere!; guaranteed to be truly random,
and if you don’t believe it, then maybe you
just don’t know what it means to be random,
‘cause you know, people aren’t very good at
being random or understanding randomness or
anything like that, so yeah.. *)
Next year, we hope to continue our experimental journey into objectivity with double-blind
submissions: by not knowing just which conference you’re submitting to, you guarantee more
fairness for everyone!
III
As if organizing SIGBOVIK wasn’t exciting enough, this year also marked the SIGBOVIK’s
first ever complete-server-downtime-during-the-seriously-final-submission-deadline. The
issue was noted around 6 pm Wasilla time, just a quarter day before submissions closed, at
which point the entirety of the SIGBOVIK Committee for Deadline Uptime excitedly jumped
into action! A flurry of email(s) was sent to Computer Club Central, the organization which
generously hosts SIGBOVIK’s servers, to little observable effect. Thanks though to the
ingeniously heroic actions of the spontaneously formed SIGBOVIK Task Force on Publicizing
the EasyChair Submission URL, submissions were almost entirely unaffected. (Whew!)
And so, without further ado, we proudly present SIGBOVIK 2010; may it enlighten and inspire!
Enjoy responsibly.
Sincerely,
The SIGBOVIK 2010 Organizing Committee
IV
Table of Contents
Track 1: Correspondence Theory
О
An Algorithm for Erdos-Bacon Number Minimization ..............................................3
Charles Garrod, Maria Klawe, Iain Matthews, Laura Trutoiu et al.
О
Classical Modal Logic: the magic of music, judgementally reconstructed .................5
William Lovas
О
The Church-Monk Isomorphism .................................................................................11
David Renshaw
О
DPS Conversion: A New Paradigm in Higher-Order Compilation .............................13
Ben Blum
Track 1: Hypertext
О
DeltaX: A New Paradigm for Machine Translation.....................................................19
Greg Hanneman
О
System Description: Google Ghost .............................................................................25
Robert J. Simmons, Brian R. Hirshman
О
The Next 700 Intuitionistic Linear Logics ..................................................................29
Lea Albaugh, William Lovas, Tom Murphy VII, Jason Reed
О
Harnessing Human Computation: β-reduction hero ....................................................33
Akiva Leffert
О
The Dissatisfactory Doily: Generating Your Own Terrible Towel Rip-Off.................41
Nels E. Beckman
О
You Keep Dying ..........................................................................................................43
Dr. Tom Murphy VII Ph.D.
О
Human Computation Method for Generating High Impact Research Papers .............45
Laura Trutoiu, Erik Zawadzki, Leigh Ann Sudol, J Marlow et al.
О
On a lexicographic ordering in spring .........................................................................51
McCall and Sons
V
Track 1: Defeating Evil
О
The Generalized Cubic Law of Nature: An Answer to Oneism Evil...........................57
Thomas Wright
О
Are you GOD? .............................................................................................................61
Anvesh Komuravelli
О
Holy States Can Save the World! ................................................................................67
Brother Jonathan Aldrich
О
Simple Systems: A Holistic, Postmodern Alternative to the yadda yadda yadda ........71
Mayank Lahiri
О
A crowdsourced approach to proving robot uprising safety ........................................75
Thomas D. LaToza
Track 1: Advanced Technology in a Magical and Revolutionary
Device at an Unbelievable Price
О
A Simplified Interface for Basic Computer Tasks .......................................................81
Jim McCann
О
T*Y*T* Fails to Solve the Halting Problem ...............................................................83
Wile Tru Do, N.R. Jaizer-Bunny
О
Developing an Algorithm for “Enhance” Functionality in Image Processing.............85
Ben Blum
О
Hyperbolic Rendering for Result Enhancement ..........................................................91
Sajme N’Ncamc
О
The Box and Circles Plot: a Tool for Research Comparison .......................................95
John Thomas Longwood
О
Department Productivity in Computer Science as a Ratio of Gender Diversity .........99
Leigh Ann Sudol
Track 1: Productivity, We Say Ironically
О
Functional Perl: Programming with Recursion Schemes in Python............................105
Robert J. Simmons, Nels E. Beckman, Dr. Tom Murphy VII, Ph.D.
О
First Year Student’s Thoughts on Reflections on Bovik’s Seminal Work: ..................111
Athula Balachandran, Richard Peng, Wolfgang Richter
VI
О
Feasibility Analysis of Cooking Spaghetti with Meatballs .........................................113
Ben Blum
О
Cataloging ....................................................................................................................115
Leigh Ann Sudol, Ciera Jaspan
О
A novel approach to managing information overload in syndicated feed readers.......119
Matt Mets
О
“Pimp My Library” Is An Affront To Pimpers Of The World, Everywhere ................121
Dr. Donna M. Malayeri, PhD
О
Computation in the Complex Time Plane....................................................................123
Matthew Rodriguez
Track 1: Perplexity Theory
О
The Reviewers Must Be Crazy: A Plea For Sanity in Peer Review ............................129
C. J. Deere
О
FAIL-Talk: Fully Automatic Incomprehensible Language Talks ................................135
Alga Rhythm
О
An Analysis and Redesign of the Evacuation Instructions in an Office Building .......139
I. B. Gone, U. Cower
О
All my friends’ researches are cooler than mine .........................................................143
Chris A. Maphone
О
How to successfully prevent the flow of information in research presentations .........145
Kami Vaniea
О
An Alternate Construction Method for the Surreal Numbers ......................................153
Jim McCann
О
The first paper on SAMIR - ANVESH ........................................................................155
SAMIR Sapra, ANVESH Komuravelli
О
The second paper on SAMIR - ANVESH ...................................................................157
О
The third and the last paper on SAMIR - ANVESH ...................................................159
VII
VIII
Track 1
Correspondence Theory
О
ach
CORRESPONDING AUTHORS
An Algorithm for Erdos-Bacon Number Minimization .................3
Charles Garrod, Maria Klawe, Iain Matthews, Laura Trutoiu,
Edilson de Aguilar, Michael Ashley-Rollman,
Harry Q. Bovik, Joanna Bresee, Jeff Carlson, Mark Desnoyer,
Michael Dinitz, Owen Durni, Jack Ferris, Ron Garcia,
Robert Harper, Benoit Hudson, Greg Humphreys,
Michelle Hutton, Kelli Ireland, Katrina Ligett, Ciera Jaspan,
Jennifer Marlow, Davin McCall, James L. McCann,
Roman Mitz, Iulian Moraru, Amar Phanishayee, David Reed,
Jason Reed, Margaret Reid-Miller, Kathryn Rivard, Or Sheffet,
Daniel D. Sleator, Ekaterina Spriggs, Matthew L. Stanton,
Mark J. Stehlik, Megan Thorsen, Vijay Vasudevan,
Kevin Waugh, Erik Zawadzki
О
THE HILLS ARE ALIVE WITH THE SOUND OF CORRESPONDENCE
Classical Modal Logic: the magic of music,
judgementally reconstructed ..........................................................5
William Lovas
О
‘ROUND CORRESPONDENCE
The Church-Monk Isomorphism....................................................11
David Renshaw
О
CORREPWNDENCE
DPS Conversion: A New Paradigm in Higher-Order
Compilation....................................................................................13
Ben Blum
1
2
An Algorithm for Erdos-Bacon Number Minimization
Charles Garrod
Pittsburgh, PA 15213
Laura Trutoiu
Harry Q. Bovik
Mark Desnoyer
Jack Ferris
Maria Klawe
Disney Research, Pittsburgh
Edilson de Aguilar
Michael Ashley-Rollman
Disney Research, Pittsburgh
Joanna Bresee
NASA Ames Research Center
Mountain View, CA 94043
Michael Dinitz
Ron Garcia
Benoı̂t Hudson
Greg Humphreys
Autodesk Canada, Montréal
Montréal, Canada
Kelli Ireland
University of Pittsburgh
Jennifer Marlow
Roman Mitz
David Reed
Creighton University
Omaha, NE 68178
Kathryn Rivard
Ekaterina Spriggs
Iain Matthews
Harvey Mudd College
Claremont, CA 91711
NVIDIA Corporation
Katrina Ligett
Cornell University
Ithaca, NY 14850
Davin McCall
La Trobe University
Melbourne, Australia
Iulian Moraru
Jason Reed
University of Pennsylvania
Philadelphia, PA 19104
Or Sheffet
Matthew L. Stanton
Megan Thorsen
Jeff Carlson
Owen Durni
Robert Harper
Michelle Hutton
The Girls Middle School
Mountain View, CA 94043
Ciera Jaspan
James L. McCann
Amar Phanishayee
Margaret Reid-Miller
Daniel D. Sleator
Mark J. Stehlik
Vijay Vasudevan
Sunnyvale, CA 94087
Kevin Waugh
Erik Zawadzki
ABSTRACT
cult. In this paper we propose Erdős Number Minimization, reducing the Erdős number of many mathematicians
and computer scientists to two. Although we do not prove
the claim, we believe that the Erdős number we obtain is op-
A small Erdős number has long been a source of pride for
mathematicians. With Paul Erdős’s death, however, obtaining an Erdős number less than two has become diffi-
3
timal without significant biological or paranormal advances.
We extend ENM with Bacon Number Reduction, a nonoptimal variant to reduce Bacon numbers, obtaining small
but non-minimal Erdős-Bacon numbers.
1.
ENM with Bacon Number Reduction (BNR), a non-optimal
algorithm to reduce Bacon numbers to obtain ENM+BNR,
and algorithm that significantly reduces Erdős-Bacon numbers. Finally, we apply ENM+BNR to obtain Erdős-Bacon
numbers no greater than six.
INTRODUCTION
6.
From his birth in 1913 to his death in 1996, Paul Erdős
co-authored nearly 1500 papers, working with nearly 500
collaborators to become the most prolific mathematician in
modern times [5]. This proliferation led mathematicians to
humorously define Erdős numbers: a person’s Erdős number
is the distance between that person and Paul Erdős in the
academic paper collaboration graph [4]. I.e., Paul Erdős
is the unique person with Erdős number zero; all of Erdős’s
immediate collaborators have Erdős number one; in general,
if you publish an academic paper with a collaborator who
has Erdős number x and none of your other collaborators
has Erdős number less than x, your Erdős number is x +
1. A similar Bacon number has been proposed for actor
Kevin Bacon, except using collaborations in movies instead
of collaborations in academic papers. Erdős-Bacon numbers
were subsequently defined to be the sum of each person’s
Erdős and Bacon numbers.
Obtaining a small Erdős number has become increasingly
difficult since Erdős’s death, but mathematicians with chutzpah have continued to publish papers with Erdős after his
death; this brings his total number of known publications to
1525, his collaborator count to 511, and the Erdős number
of the chutzpah-bearing mathematician to one. The latest
publications co-written with Paul Erdős appeared more than
ten years after his death, and with additional rumored works
in progress, Erdős’s publication list is expected to grow.
In this paper, we describe a technique called Erdős Number Minimization (ENM). Although we do not prove the
claim, we believe that, without significant biological or paranormal advances, ENM is optimal for mathematicians who
have been unable to work with Erdős in the past. We supplement ENM with a similar but non-optimal Bacon Number
Reduction algorithm to obtain ENM+BNR.
2.
7.
ERDOS NUMBER MINIMIZATION
BACON NUMBER REDUCTION
See [6] for more information.
4.
EVALUATION OF ENM+BNR
In [3], Maria Klawe co-authored with Paul Erdős, thus
guaranteeing that all authors on this paper have an Erdős
number no greater than two. Additionally, Iain Matthews
appeared in [2] with Sigourney Weaver, who appeared in
[1] with Harry Connick Jr., who appeared in [7] with Kevin
Bacon. By [6], all collaborators on this paper also appear
in the movie credits about the paper and thus have a Bacon number no greater than four, making their Erdős-Bacon
number no greater than six.
5.
REFERENCES
[1] Ann Biderman and David Madsen. Copycat, 1995.
http://www.imdb.com/title/tt0112722/.
[2] James Cameron. Avatar, 2009.
[3] Paul Erdős, Frank Harary, and Maria Klawe. Residually
complete graphs. In Proc. Sympos. Combinatorial
Mathematics and Optimal Design, pages 117–123,
Colorado State University, Fort Collins, Colorado, 1978.
[4] Caspar Goffman. What is your Erdős number? The
American Mathematical Monthly, 76(7):791, 1969.
[5] Jerry Grossman, Patrick Ion, and Rodrigo De Castro.
Erdős number project, 2010.
http://www.oakland.edu/enp/.
[6] Iain Matthews, Laura Trutoiu, Charles Garrod, and
Maria Klawe et al. Bacon Number Reduction, 2010.
http://tinyurl.com/ENM+BNR.
[7] Willie Morris and Gail Gilchriest. My Dog Skip, 2000.
2.
3.
ACKNOWLEDGMENTS
This paper was most definitely not authored by Luis von
Ahn, David G. Andersen, Alexei Efros, Michelle Goodstein,
Raffay Hamid, Mike Mistry, Don Sheehy, Robert Simmons,
or Ronit Slyper.
CONCLUSIONS
In this paper, we propose Erdős Number Minimization
(ENM), an algorithm to minimize Erdős numbers for authors unable to collaborate with Paul Erdős. We extend
4
Classical Modal Logic
the magic of music,
judgementally reconstructed
William Lovas
March 21, 2010
Abstract
♩ The hills are alive... with the sound of music. 1
Introduction
Typically, musicians concern themselves with just one or two notions of music,
in the absolute. Modal music is a generalization of traditional music in which we
can consider music from multiple perspectives, or modes. In fact, modal music
dates back to the antiquities, and so we may more properly speak of classical
modal music. In what follows, we develop a logic for classical modal music based
around a glorious and perfect septuality of seven different modal judgements.
2
The Logic of Classical Modes: Judgements
Ordinary intuitionistic logic is typically concerned with the judgement A true,
meaning that the proposition A is true. Classical modal logic is a multimodal
logic [Ree09] based on seven judgements, each one representing one of the classical modes:
• C ionian — C has the Ionian (or “major”) mode
• D dorian — D has the Dorian mode
• E phrygian — E has the Phrygian mode
• F lydian — F has the Lydian mode
• G mixolydian — G has the Mixolydian mode
• A aeolian — A has the Aeolian (or “minor”) mode
• B locrian — B has the Locrian mode
5
Γ Δ
Θ Λ Ξ Π
Σ Φ Ψ Ω
Figure 1: Greek letters appropriate for use as logical contexts.
Since each of these judgements may appear on both the left- and the righthand side of a sequent, we split the usual context into seven zones, one for each
modal judgement. A difficulty presents itself in choosing appropriate names for
the contexts, as there are only 10 capital Greek letters that are not written
identically as some Roman capital letter1 (see Figure 2). We make the obvious
choices for the Dorian (Δ), Phrygian (Φ), and Lydian (Λ) contexts. Mixolydian
with its infixed x gets as its context the Greek letter Ξ. The usual ordinary or
“major” mode, Ionian, gets the usual ordinary context, Γ. The Aeolian or
“minor” mode gets Σ, which looks kind of like a sideways “M”. And finally, the
highly unstable and notoriously difficult-to-resolve Locrian mode is represented
by Ω, a symbol which calls to mind non-termination and inconsistency. The
three remaining unused Greek letters (Θ, Π, and Ψ) are left to future work.
Classical modal logic thus involves sequents of the form:
Γ; Δ; Φ; Λ; Ξ; Σ; Ω J
for J one of the seven modal judgements, and where
Γ = C1 ionian, . . . , Ci ionian
Δ = D1 dorian, . . . , Dj dorian
Φ = E1 phrygian, . . . , Ek phrygian
Λ = F1 lydian, . . . , Fl lydian
Ξ = G1 mixolydian, . . . , Gm mixolydian
Σ = A1 aeolian, . . . , An aeolian
Ω = B1 locrian, . . . , Bλ locrian
This excessively complex sequent is in keeping with the classical nature of the
logic: classical music is serious business!
We want to be able to make inferences like:
Γ; Δ; Φ; Λ; Ξ; Σ; Ω C ionian
Γ; Δ; Φ; Λ; Ξ; Σ; Ω D dorian
Therefore, in addition to the usual hypothetical reasoning principles, we have
judgemental principles corresponding to the glorious septuality.
Principle 1 (Septuality) For all notes N , modes m, and natural numbers k:
Γ; Δ; Φ; Λ; Ξ; Σ; Ω N m if and only if Γ; Δ; Φ; Λ; Ξ; Σ; Ω (N +7 k) (m +7 k).
where the transposition operator +7 is defined in the obviously canonical way.
1 excluding the obsolete letters digamma, qoppa, san, and sampi, most of which are unavailable in LATEX.
6
Figure 2: The Ciiiiiiiiircle... the Circle of Fifths! 3
Circle of Fifths
The Principle of Septuality hints at the deep symmetries of classical modal
music, brought out by the Circle of Fifths (Figure 2). To internalize the notion
of transposition, we introduce the connective N , read “transpose N ”.
Γ; Δ; Φ; Λ; Ξ; Σ; Ω N (m +7 1)
Γ; Δ; Φ; Λ; Ξ; Σ; Ω N m
Γ; Δ; Φ; Λ; Ξ; Σ; Ω N m
I
Γ; Δ; Φ; Λ; Ξ; Σ; Ω N (m +7 1)
E
But of course, a circle goes both ways, so as is traditional in modal logics, we
also have the backward-looking dual connective N , read “untranspose N ”.
Γ; Δ; Φ; Λ; Ξ; Σ; Ω N (m −7 1)
Γ; Δ; Φ; Λ; Ξ; Σ; Ω N m
Γ; Δ; Φ; Λ; Ξ; Σ; Ω N m
I
Γ; Δ; Φ; Λ; Ξ; Σ; Ω N (m −7 1)
E
where, again, the untransposition operator −7 is completely standard.
The rules for both connectives are easily seen to be locally sound and complete. But more interestingly, the connectives lend themselves easily to a clear
axiomatic characterization. The first axiom internalizes the Principle of Septuality,
N ≡ N
that is to say, is 7-involutive. The second and third axioms capture the notion
that transposition and untransposition are mutually inverse,
N ≡ N ≡ N
7
4
4.1
Future Directions
Relationships with Other Logics
It is well-known that the intuitionistic modal logic S4 [PD01] has exactly seven
distinct generalized modalities: ·, , , , , , and . An obvious
direction of future work would be to connect these seven modalities to the seven
modal judgements of our classical modal music logic.
4.2
Other Connectives
Generalizing from our experience with modern music, we might infer the existence of connectives N and N , some instances of which would have interpretations “between the modes”, with the following axiomatic characterizations:
(N ) ≡ N
(N ) ≡ N
along with rules like
N phrygian
N lydian
An obvious question to ask is just what modal judgemental principles would
these connectives correspond to? How could they be “externalized” as judgements in a globally sound and complete way?
Of course, for algebraic completeness, we would want to include the identity
or “natural” connective N , where
N ≡ N
which derives semantically from the well-known categorical notion:
η -
G
F
F♩
?
F
G♩
η
8
?
G
4.3
Proof Theory
A rich and expressive proof theory cries out to be invented2 in which proofs are
compositions and proof checking is harmoniousness validation. Naturally, this
proof theory should be amenable to canonical forms [WCPW02, Bac47c].
4.4
Classical Classical Modal Logic
In this paper we cover the intuitionistic variant of classical modal logic, but
one could easily consider a multiple-conclusion style classical variant. The chief
difficulty is coming up with enough Greek letters to round out the zones, since
as explained in Section 2, the present work leaves only 3 suitable letters. One
rather pedestrian option would be to leverage subscripts:
ΓL ; ΔL ; ΦL ; ΛL ; ΞL ; ΣL ; ΩL =⇒ ΓR ; ΔR ; ΦR ; ΛR ; ΞR ; ΣR ; ΩR
(Yawn.) A more creative option would be to press into service the 4 canonical
obsolete Greek letters, digamma, qoppa, san, and sampi, bringing the total
unused letters to a perfect 7 — a tempting coincidence! But unfortunately,
these letters are difficult to typeset, and in any case bear little connection to
the logical content of the seven zones.
So a cleverer alternative would be to use very powerful typesetting features to
allow us to reuse our carefully chosen seven contexts while clearly distinguishing
the “false” ones from the “true” ones:
Γ; Δ; Φ; Λ; Ξ; Σ; Ω
Γ; Δ; Φ; Λ; Ξ; Σ; Ω An advantage of this notation is that sequents may be read either right-side up
or upside-down, left-to-right or right-to-left, a trick used to great comic effect by
musical artists such as Borge [Bor90] and Bach [Bac47b, Bac47a] and computer
scientists such as Hofstadter [Hof79] and Mairson [Mai05].
Finally, a third option would be to follow in the footsteps of many presentations of classical linear logic [Gir87], which leverage duality to halve the number
of required zones, and find some way of leveraging septuality, a concept seven
times more powerful, to shrink the size of a classical modal sequent.
5
Coda
If modals be the logic of love, calculemus!
2 “Discovered.”
— Ruy Ley-Wild
9
References
[Bac47a]
Johann Sebastian Bach. Canon 1, a 2 cancrizans (“crab canon”).
In The Musical Offering [Bac47c].
[Bac47b]
Johann Sebastian Bach. Canon 9, Quærendo Invenietis. In The
Musical Offering [Bac47c].
[Bac47c]
Johann Sebastian Bach. The Musical Offering. BWV 1079. July
1747.
[Bor90]
Victor Borge. William Tell backwards. In The Best of Victor
Borge: Act One & Two. GMZ Productions, 1990. Available from
http://www.youtube.com/watch?v=MuWUp1M-vuM.
[Gir87]
Jean-Yves Girard. Linear logic. Theoretical Computer Science,
50:1–102, 1987.
[Hof79]
Douglas R. Hofstadter. Crab canon. In Gödel, Escher, Bach:
an Eternal Golden Braid, pages 199–203. Basic Books, New York,
1979.
[Mai05]
Harry Mairson. Constructive classical logic. POP Seminar at
Carnegie Mellon University (http://www.cs.cmu.edu/afs/.cs.
cmu.edu/Web/Groups/pop/seminar/050315.html), March 2005.
[PD01]
Frank Pfenning and Rowan Davies. A judgmental reconstruction of
modal logic. Mathematical Structures in Computer Science, 11:511–
540, 2001. Notes to an invited talk at the Workshop on Intuitionistic Modal Logics and Applications (IMLA’99), Trento, Italy, July
1999.
[Ree09]
Jason Reed. A judgmental deconstruction of modal logic. Available from http://www.cs.cmu.edu/∼jcreed/papers/jdml.pdf,
January 2009. Revised and redacted at http://www.cs.cmu.edu/
∼jcreed/papers/jdml2.pdf, May 2009.
[WCPW02] Kevin Watkins, Iliano Cervesato, Frank Pfenning, and David
Walker. A concurrent logical framework I: Judgments and properties. Technical Report CMU-CS-02-101, Department of Computer
Science, Carnegie Mellon University, 2002. Revised May 2003.
10
The Church-Monk Isomorphism
David Renshaw
“All musicians are subconsciously mathematicians.” —Thelonious Monk
“And vice versa.” —Alonzo Church
Church’s Thesis:
(∀x.∃y.φ(x, y)) → (∃f.∀x.∃y u.T(f, x, y, u) ∧ φ(x, y))
Monk’s Mood:
References
[1] W. Lovas. Classical Modal Logic: the magic of music, judgementally
reconstructed. SIGBOVIK 2010.
[2] D. Renshaw. The Davis-Davies Isomorphism. To appear (maybe).
11
ach
sigbovik
Plenary Program Committee
2009 Confidential Paper Reviews
Paper 3: The Church-Monk Isomorphism
Reviewer 1: Jeffrey Renshaw
Rating: 3 (strong accept)
Confidence: 3/4
What a remarkable coincidence that I was asked to review this paper! Just the other day I was pondering some deep connections between Karl Marx and Karl Goldmark. David’s paper brilliantly elucidates
many ideas that I had been struggling to grasp. It looks like we’re on to something big here. Apparently, isomorphisms are far more common than was previously believed. I eagerly anticipate future work
in this direction.
12
DPS Conversion: A New Paradigm in Higher-Order
Compilation
Ben Blum (bblum@andrew.cmu.edu)
2010.04.01
1
Abstract
This paper introduces a new judgment and accompanying system of rules for use in some higher-order
compilation stages that allows reduction of noobs into more damaged noobs, and finally conversion into gibs.
The translation is done by repeated applications of damage-per-second (DPS), which we denote dps. The
primary judgment is Γ, dps noob ↓ gibs, which is mostly useful only in first-person shooter or MMO cases.
We provide a proof for local soundness and completeness, and finally an example usage case.
2
Definitions
In certain situations, we come across constructs of the following form that we wish to DPS-convert.
noob ::= ×n [hp : int, . . . ]
dps ::= ROCKET | AWP | SPELLCASTER | SWORD | . . . | HEADSHOT[dps] | × [dps, dps]
gibs ::=
A
• The rest of the components of noob are system-specific. All we care about is the noob’s hp total.
• Note that the final construct in the dps definition represents that damage can be dealt from multiple
sources at a time, and we are only concerned about the total.
3
Rules
We set forth a series of rules for performing the following conversion:
dps a
Γ, dps
Γ, dps
int
(1)
(2)
noob noob
noob ↓ gibs
(3)
We assume standard arithmetic operations +, −, and ∗, and let Γ = | Γ, dps.
1. The a judgment tells us how much damage a given dps does. The exact implementation of these
rules may vary by system; we present an example.
ROCKETa25
AWPa15
SPELLCASTERa8
δaα
boom.headshot
HEADSHOT[δ]a3 ∗ α
13
SWORDa10
δ1 aα1 δ2 aα2
multiple.dps
×[δ1 , δ2 ]aα1 + α2
2. This rule tells when the amount of dps presently given is not sufficient for reducing the noob to gibs.
Applications of this rule are guided by how much dps is in the context.
Γ ν ν δaα π1 ν − α > 0
not.dead.yet
Γ, δ ν ×n [π1 ν − α, π2 ν , . . . , πn ν ]
νν
insufficient.dakka
3. This judgment represents the end case in DPS conversion. These rules are similarly inductive on the
context.
Γν↓
Γ ν ν δaα π1 ν − α ≤ 0
lol.pwned
overkill
Γ, δ ν ↓
Γ, δ ν ↓
A
A
A
3.1
Soundness
Each class of dps obviously makes some sound, proportional to the amount of damage it does. For example,
using ROCKET obviously makes a loud explosion, while a HEADSHOT would cause the player to shout out in
excitement. Therefore, to ensure soundness, you should do the most DPS possible.
3.2
Completeness
As a wise man once said [1], “Any problem is solvable with sufficient firepower.” This serves as an inductive
proof that every noob will eventually be reduced to gibs with enough dps.
4
Future Work
The DPS Conversion system as presented only cares about the hp total of noobs. A more advanced system
would be able to handle healers and tanks. In the former case, certain abilities could cause the noob’s hp
to increase in addition to decreasing during a derivation. The latter case makes relevant the other elements
of the noob tuple - in addition to hp, a noob might perhaps have spell resistance which would negate
DPS from a SPELLCASTER. A new version of DPS would be able to generalize such abilities and include their
effects in the conversion process.
5
Conclusion
We provide an example DPS derivation. For simplicity, let noob ::= ×1 [hp]. Our noob will start with 50hp.
In the first second, a SPELLCASTER gets a HEADSHOT; in the next, a ROCKET and SWORD join the battle, and
finally the AWP shoots, though the noob was successfully converted to gibs in the first two seconds.
SPELLa8
insufficient.dakka
boom.headshot
×1 [50] ×1 [50]
HS[SPELL]a24
not.dead.yet
HS[SPELL] ×1 [50] ×1 [26]
HS[SPELL], ×[ROCKET, SWORD] ×1 [50] ↓
A
HS[SPELL], ×[ROCKET, SWORD], AWP ×1 [50] ↓
ROCKETa25 SWORDa10
×[ROCKET, SWORD]a35
multiple.dps
lol.pwned
A overkill
Acknowledgments: Special thanks to Joshua Wise for the insufficient.dakka rule, Nicholas Tan1 for the
future work suggestions, and Zachary Z. Sparks for the soundness proof.
6
References
1. Wiseman, A. (22 B.C.) Any problem is solvable with sufficient firepower. On archaeologically preserved
stone tablets.
14
ach
sigbovik
Paper 11: DPS Conversion: A New Paradigm in
Higher-Order Compilation
Reviewer 1
Rating: 2 (accept)
Confidence: 1/4
The author formulates a sound and complete operational semantics for the reduction of noobs to gibs.
While this process is of considerable commercial interest, it has until now been implemented in a rather
ad-hoc way, so the paper makes a strong contribution by putting it on firm theoretical footing.
Particular strengths of the paper include its modular proof of soundness, which extends without change
to any sound-producing attack (however, this approach does not apply to silent-but-deadly weapons);
and also its use of the spooky skull-and-crossbones glyph.
The author could improve the paper by generalizing the result beyond first-person shooters; the extension to person-shooters of other ordinals seems straightforward. Also, I am fairly sure that there is
relevant prior workwhich the author has not cited---if I am not wrong Murphy’s work on Generalized
Super Mario Bros. and Beckman’s on Infinite Arkanoid are pertinent.
While I have acquired, over the years, certain knowledge of gamer culture, my confidence rating is low
because on the whole I am more of a Scrabble sort of guy. Possibly by continuing to review 3L33T
G4M3RZ type of papers I will be able to “level up” my confidence in this area. I suppose I couldalso
hire them out to a “review farmer” and save myself some carpal tunnel damage.
15
16
Track 1
Hypertext
О
ach
HYPERTEXT TRANSLATION
DELTAX: A New Paradigm for Machine Translation ......................19
Greg Hanneman
О
HYPERTEXT TRANSFER PROTOCOL DOWNLOAD FREE
System Description: Google Ghost................................................25
Robert J. Simmons, Brian R. Hirshman
О
HILL ALSO STANDS FOR “HYPERTEXT INTUITIONISTIC LINEAR LOGIC”
The Next 700 Intuitionistic Linear Logics .....................................29
Lea Albaugh, William Lovas, Tom Murphy VII, Jason Reed
О
HYPERLAMBDA
Harnessing Human Computation: β-REDUCTION HERO ....................33
Akiva Leffert
О
THE HUMID HYPERTEXT!
The Dissatisfactory Doily: Generating Your Own Terrible Towel
Rip-Off ...........................................................................................41
Nels E. Beckman
О
YOU HYPERTEXT AND YOU KEEP DYING
You Keep Dying.............................................................................43
17
(continued on next page)
Track 1
Hypertext (continued)
О
ach
HYPERIMPACT
Human Computation Method For Generating High Impact
Research Papers .............................................................................45
Laura Trutoiu, Erik Zawadzki, Leigh Ann Sudol, J Marlow,
Anonymous Coward, Ekaterina Taralova, Field Cady,
Matt Stanton, Ciera Jaspan, Or Sheffet
О
HYPERORDER
On a lexicographic ordering in spring ...........................................51
McCall and Sons, Semantic Exporters
18
DeltaX: A New Paradigm for Machine Translation
Greg Hanneman
Language Enthusiast and Errant Scribbler
1
Introduction
2
Machine Translation
The problem of machine translation is usually expressed as the task of automatically translating text
(e.g. a sentence) in one human language (e.g. English) into another human language (e.g. Azerbaijani), such that the semantic meaning of the original
sentence is also expressed by the translation. The
precise formulation of the task, however, has varied
from paradigm to paradigm. Because the mathematical models in these things quickly get complicated,
researchers who want their work to be understood
by others frequently take recourse in a generative
As you may suspect from the way that last para- story: a simplified conceptual overview of how their
graph ended, in this paper we introduce just such a systems translate sentences.
breakthrough. We describe DeltaX, which we believe to be a completely novel paradigm for carrying out machine translation. In contrast to previous 2.1 A Generative Story
methods, DeltaX is fast, cheap, easy, and requires Once upon a time, there was a man called Warren
minimal additional hardware beyond what may easily Weaver. Mr. Weaver was a mathematician, which
be obtained by ordinary consumer-level technology means that his job was to solve math problems all
users.
day. This was a very hard job back when a computer
The remainder of this paper is organized as follows. filled an entire room! But Mr. Weaver was a very
Since we have no set page-count limit, we first skip a smart man, and so he enjoyed his job.
One day, Mr. Weaver was walking to his office
whole bunch of lines just to thumb our nose at other
conferences. Next, in Section 2 we briefly review the when he thought of a fun new math problem to solve.
“When a man speaks Russian,” Mr. Weaver thought,
“what that really means is that he’s speaking English.” (Mr. Weaver had a very good imagination.)
“He speaks English in his head, but then the English
goes through a secret code and comes out as Russian
instead.” Mr. Weaver thought very hard. “Now, all
scientific question of machine translation and sum- I have to do figure out what that secret code is, and
marize its existing paradigms. Section 3 is our mag- then I can translate the Russian back to English!”
num opus1 , where we describe DeltaX in detail. We Do you remember hearing about secret decoder rings
follow with Section 4, which performs a comparative on the radio, where one letter stands for another?
analysis between DeltaX and other machine trans- Well, all Mr. Weaver had to do was figure out which
Russian words stood for which English words in the
lation methods. Finally, Section 5 concludes.
secret Russian code, and then he would know just like
that how to translate Russian.
1 Op. 4, in E minor
Mr. Weaver went straight to work. He worked very
The field of machine translation has experienced considerable growth since its original conception in the
late 1940s. Over the past 60 years, countless countably finite researchers have developed new techniques
and methodologies to advance the state of the art
(and the state of their CVs). Sometimes these advancements are significant enough that it can be
claimed that a new paradigm of machine translation has been developed, though to date there have
been very few such breakthroughs.
19
hard for many days, but the secret code was also
very hard. Mr. Weaver was never able to figure it
out. He told many other people his idea, but they
all laughed at him. At last he stopped working and
died. But his story has a happy ending. Even though
Mr. Weaver did not find out the secret Russian code,
he invented the new problem of statistical machine
translation. The end.2
at this crap: Can you can fruit in that can? Three
different parts of speech! Then you gotta put all the
words together, but if the man saw the boy with a
telescope, how am I supposed to know who’s got the
looking glass! Man, this is so unfair! Dad said I
gotta stay in my room until I finish parse trees for the
garden path sentences, but I think I’m gonna sneak
out. There are just too many rules to write down!
2.2
2.3
Give Yourself Goosebumps
Chapter 1. Hi! My name is Rule-Based Machine
Translation, but you can call me Rules for short. It
totally rules! Anyway, I gotta tell you about my old
man — he’s such a pain in the neck! I just know that,
if you met him on the street, you’d run away screaming. All he talks about is his stupid Russian decoder
rings, which is so old-fashioned. Can you imagine!
Fiddling around all day with his stupid mathematical rules, as if all the wicked cool things about human
language could be expressed as mere numbers.
Let me tell you how the modern kids see things
these days. We like rules, man, ’cause they totally
rule! Just check this out. You know how in French
you’ve got thinks like la voiture bleue, right, where
the adjective comes after the noun it modifies? But
when you say the same thing in English, you’ve totally gotta reverse the order so the adjective comes
first – the blue car? Well, that’s the whole secret! If
we want to translate French into English, we’ve just
gotta know the rules of the game. No numbers, no
statistics.
Chapter 2. Now I gotta tell you the best part:
learning rules is a whole lot easier than breaking some
made-up secret code. We just have to ask people
what they are! I mean, you speak English, so you
can just tell me what order the words in a sentence
have to come in, and then we’ll figure out what categories they belong to, and then someone who knows
French’ll tell us the same information there. We put
it all together and, wham! That’s machine translation.
Chapter 3. Aw, man, I just got sent to my room
by my dad ’cause my homework was late. It was
so stupid! All I had to do was figure out a set of
rules that would explain all languages, which is like
totally reasonable, right? That’s what I thought until
I started my homework. Do you know how many
different meaning there are for some words? Look
2 Adults reading this paper in the presence of small children
should take care at this point to show the pictures.
All Grown Up
Good evening, everyone, and thank you for coming
to my presentation. Tonight’s topic is going to be
phrase-based statistical machine translation,
and I would like to say right at the beginning how
pleased I am to be able to speak to you all about it
at such an important conference.
Now, as we all know, the field of machine translation has faced some significant problems in the past.
We spent a long time paying linguists to write thousands of synchronous grammar rules, morphological
analyzers, parsed treebanks, and who knows what
else, and at the end of the day none of those techniques were any help. Those processes are just too
expensive, and getting good results out of them just
takes too long. Who cares whether a word’s in the
subjuctive aorist passive, or if Arabic is a VSO language? Every time I fire a linguist, my team’s translation results go up five points!
Friends, the real answer is data. Data, data, data.
We’ve developed cutting-edge mathematical models
to automatically extract information from parallel
text — and it’s better information than anything a
room full of linguists will tell you. In this slide, you
can clearly see the way to do translation. Just get
a whole bunch of sentences in language F here, and
match them up with their translations in language E
here. International parliaments are great sources of
these, ready-made! Line up all your sentences, then
run statistical word alignment to let the math tell you
exactly which words in F translate to which words in
E exactly which percentage of the time. No made-up
linguistics — the numbers don’t lie.
On this slide you can see the next step. Once you
know how the words translate, just group a bunch
of words together, and you know how the phrases
translate too. Question, ma’am? No, no linguistics
here either. When I say “phrase,” I just mean any old
group of contiguous words in F that translates into a
group of contiguous words in E. Again you can rely
on the mathematical models we’ve developed to give
you accurate translation probabilities.
20
Finally, here on Slide 3, you see how easy it is to
translate. When you get a new sentence in F , just
look through your scored phrase pairs for things that
match bits and pieces of the F sentence. String ’em
all out in a line, multiply their probabilities, and take
the ones that give you the highest score in E. Whatever the data supports the most is the translation you
give: let Bayes’ Rule do all the work for you, and just
maximize that old P (E | F ).
To conclude, thank you all again for coming
tonight, and remember: there’s no data like more
data!
3
3.1
• “Motion of a body in which every point of the
body moves parallel to and the same distance as
every other point of the body”
A New Paradigm: DeltaX
f(a)
Motivation
In our preliminary work on DeltaX, we were primarily struck by the numerous shortcomings in existing machine translation paradigms. In informationtheoretic statistica machine translation, for example,
we found ourselves dissatisfied with the reduction of
the infinite beauty and creativity of natural language
to a mere case of cipher-cracking. On the other hand,
the intensive manual effort required to create precise
grammars and lexicons in rule-based machine translation leaves us searching for a faster, more automatic
method. Finally, we believe that the data-driven
techniques of phrase-based statistical translation are
reaching a saturation point: while they work well for
languages with very large amounts of parallel text,
they perform very poorly in the absence of these great
resources.
Ideal machine translation has three qualities: (1)
it is fully automatic, (2) it is high-quality, and (3)
it handles a broad coverage of text styles and domains. It has been observed that, at the present time,
one may apply the existing paradigms and obtain any
two of these three criteria. This, we believe, is not
enough.
3.2
out the distracting probabilistic models or formal
grammars of previous approaches. The result is fast,
cheap, easy, and requires minimal additional hardware beyond what may easily be obtained by ordinary
consumer-level technology users.3
Fundamentally, the definition of translation is
clear:
The DeltaX Formulation
a'
a
Figure 1: Simple formulation of the translation task.
This is exemplefied in Figure 1. Backing away
from morphology, grammars, and parallel data, we
frame the task of translation simply: its goal is to
find a function f (a) that maps a given point a onto
the point a . Since our starting and ending points
are both known, we can easily define this function
in terms of the distance between a and a . We further use the fact that we can re-write point a as the
ordered pair (x, y).
f (a) =
f (x, y) =
a − a
(x , y ) − (x, y)
(1)
(2)
y − y
f (x)
=
=
0
x − x
(3)
(4)
f (x)
=
Δx
(5)
We thus arrive at our function for translation. To
address the machine portion, we apply the examplebased defintion of “machine” as “computer,” as seen
in common locutions such as “Linux machine” or
“machine room.” Putting the two halves together
provies a clean, intuitive new paradigm for machine
translation, which we fittingly call DeltaX.
We believe that the current limitations of machine
translation technology in all existing paradigms come
from the fact that the problem is stated incorrectly in
all of them. Recall that we previously introduced machine translation as the task of automatically translating a sentence in one human language to a sen3 Did you notice we stole that sentence almost word for word
tence in another human language. In DeltaX, we
return to this fundamental and clarifying truth with- from the introduction? All the good papers do it.
21
le Dell
Dell
9.84'
3m
7m
Figure 2: Setup for our English–French machine translation experiments. Our Dell Optiplex input is translated approximately a net 7 meters.
Carrying out the machine translation task with
DeltaX is simple, requiring only an input, a goal,
and sufficient force — all readily obtainable by today’s underfunded and small research groups. For the
particularly miserly, the force required may even be
calculated in advance with the application of simple
physics. We now describe each of the requirements
in more detail.
• Input. Borrowing from the classical machine
translation literature, the input is typically
termed the source.4 As a general translation
platform, DeltaX places no restriction on the
source: it can be any machine — running in any
language — the user prefers.
• Goal. Classical machine translation uses the
term target to refer to the output of a translation. Here, we allow the target to also be generic:
the user must only pick a goal location that the
source machine should be translated to.
• Force. Typically, force for a machine translation system is provided by an amount of computing power — either that or an impending
deadline, frustrated advisor, or project evaluations. However, many modern systems require
large computing clusters, fancy servers, or both.
4 Actually, in the Late Classical period, P. M. T. Bach wrote
it ye olde Sor e, but we must make allowances for modern
spelling and typography.
In DeltaX, we relax this requirement, specifying only the use of some amount of motive power.
The exact amount may be calculated exactly and
in advance.5
4
Experiments
To test our system, we carried out experiments on
English-to-French machine translation. In each case,
our source was a Dell Optiplex GX260. The target was a point in the next machine room over, approximately 15 meters away, and the force was two
minimum-wage undergraduates in reasonable physical condition. This setup is shown in Figure 2.
A common metric in machine translation is the
BLEU score, a geometric mean of n-gram translation precision compared to a human-produced reference translation.6 Since our system uses neither
n-grams nor human-produced references, we find
BLEU scores inadequate for our evaluations. Instead,
we use an adapted metric that we term CHEESE,
or “Computer Here Evaluation, Each System Expected.” This easy-to-use metric is based on the
target specified during the translation task — if the
end result is that the machine has been translated to
5 This is the second time we have made this statement. The
proof and derivation, as in all such cases, are left as an exercise
to the reader.
6 If you didn’t understand that last sentence, good luck with
“real” research papers. They’re all like that!
22
the target position, CHEESE returns success. If not,
CHEESE returns failure status and fires the undergraduate movers (“human translators”).
We compare DeltaX to existing machine translation systems freely available on the Internet. Google
Translate is a statistical system built from approximately 4.3 quintagazillion words of text in all domains and genres, including some that writers haven’t
invented yet. Systran (or “Babelfish” or “AltaVista
Translate,” depending on your age) is a rule-based
system made up of context-free grammar rules, begun in 1961 and currently scheduled for completion
by human linguists in 2025. To make our comparison fair, we use the same input for all systems. The
Dell Optiplex used in the DeltaX system is first converted into a pure text representation by serializing
its instruction manual, Verilog files for its processors,
and contents of its hard drive. We next pay human
translators (in the classical sense) to translate the
resulting text into French; this provides a reference
translation for use with the BLEU metric.
System
Google Translate
Systran
DeltaX
Metric
0.0442
0.0657
1.0000
Score
BLEU
BLEU
CHEESE
Figure 3: Results of experiments comparing the
DeltaX machine translation system to two Internet
competitors.
Results are shown in Figure 3. Surprisingly, the
rule-based system outperforms the statistical system
in our machine domain. We hypothesise that this
is true because machine translation — as we have
expressed it — can be succinctly expressed a series of
generally applicable rules, something that rule-based
systems excel at. Neither of the two Internet-based
machine translations, however, hit upon the correct
series of rules to apply in our experimental setup.
The DeltaX system, on the other hand, was able to
express the translation task as a series of three rule
applications:
1. Never talk about machine translation.
2. Translate south by three meters.
3. Translate east by seven meters.
4. Translate north by three meters.
By following this simple procedure, the DeltaX system is able to achieve a perfect 1.0 on the CHEESE
metric, while the competing systems fail to measure
up in BLEU.
5
Conclusion
.sedulcnoc 5 noitceS ,yllaniF .sdohtem noitalsnart
enihcam rehto dna XatleD neewteb sisylana evitarapmoc a smrofrep hcihw ,4 noitceS htiw wollof eW
.liated ni XatleD ebircsed ew erehw ,supo mungam
ruo si 3 noitceS .smgidarap gnitsixe sti ezirammus
dna noitalsnart enihcam fo noitseuq cifitneics eht
weiver ylfeirb ew 2 noitceS ni ,txeN .secnerefnoc rehto ta eson ruo bmuht ot tsuj senil fo hcnub elohw a
piks tsrif ew ,timil tnuoc-egap tes on evah ew ecniS
.swollof sa dezinagro si repap siht fo redniamer ehT
.sresu ygolonhcet level-remusnoc yranidro yb deniatbo eb ylisae yam tahw dnoyeb erawdrah lanoitidda
laminim seriuqer dna ,ysae ,paehc ,tsaf si XatleD
,sdohtem suoiverp ot tsartnoc nI .noitalsnart enihcam tuo gniyrrac rof mgidarap levon yletelpmoc
a eb ot eveileb ew hcihw ,XatleD ebircsed eW
.hguorhtkaerb a hcus tsuj ecudortni ew repap siht ni
,dedne hpargarap tsal taht yaw eht morf tcepsus yam
uoy sA
.shguorhtkaerb hcus wef yrev neeb evah ereht etad
ot hguoht ,depoleved neeb sah noitalsnart enihcam fo
mgidarap wen a taht demialc eb nac ti taht hguone
tnacifingis era stnemecnavda eseht semitemoS .)sVC
rieht fo etats eht dna( tra eht fo etats eht ecnavda
ot seigolodohtem dna seuqinhcet wen depoleved evah
srehcraeser etinif ylbatnuoc sseltnuoc ,sraey 06 tsap
eht revO .s0491 etal eht ni noitpecnoc lanigiro sti
ecnis htworg elbaredisnoc decneirepxe sah noitalsnart
enihcam fo dleif ehT
Well, even high school students know that the conclusion section is just the introduction section written
backwards.
Acknowledgements
The authors — even though there is only one of us
— would like to thank Lea Albaugh (“B-A-U-G-H,
Pennsylvania Dutch–style”) for her fine figure drawing contributions to the present paper. The research
reported here was supported by borrowed time from
the Language Tecnologies Institute at Carnegie Mellon University when the author was supposed to be
writing his thesis proposal. Conclusions drawn from
23
it represent the author’s own artwork and do not represent the opinions of real artists who actually know
how to use charcoal sticks and pencils.
References
[1]
[2]
[3]
Perl
C/C++
Java
\$scalar
int* numKumquats
Object o
24
System Description: Google Ghost
Robert J. Simmons and Brian R. Hirshman
Department of Computer Science
{rjsimmon,hirshman}@cs.cmu.edu
Abstract
This paper describes the new craze sweeping through computer science
departments located along Forbes Avenue in Pittsburgh, PA, USA: playing Google Ghost. This paper describes the static and operational semantics of Google Ghost, and provides a constructive proof that you1
can become a Google Scholar!
1
Introduction & Motivation
Google Ghost is a brilliant new adaptation of the deprecated, archaic children’s game
Ghost. A wise horde once said:
Ghost is a word game in which players take turns adding letters to a
growing word fragment, trying not to be the one to complete a valid
word. If a player completes a word, they lose that round of the game and
starts a new round. Each fragment must be the beginning of an actual
word. Usually some minimum is set on the length of a word that counts,
such as three or four letters.2
We identify several fundamental problems with the deprechaic version of Ghost:
• Boredom: Ghost is boring as crap.
• Difficulty: Oh, are you done with your New York Times crossword puzzle, nerd?
• Speed: Ghost tends to require lots of thought before a turn is played, as the player
tries to both guess the word the other players were trying to spell and try to (potentially) subvert their intent by redirecting to another word without completing
the original word.
• Size: The creative possibilities of Ghost are limited by the people in the room with
you. And good luck getting people to play your boring, hard, slow game, nerd.
These problems are all solved by Google Ghost, which is awesome.
1
2
Yes, you.
http://en.wikipedia.org/wiki/Ghost (game)
25
2
Google Ghost
Google Ghost is a multiplayer game that is played by using, instead of boring English
words, the massive database of mapreduced by Google Suggest.3 To play Google Ghost,
you need two or more players and one ghostwriter4 who has access to a computer running
a modern, JavaScript-enabled web browser. Play starts with the player to the right of the
previous loser utters a word (either an English word or common, proper noun) and then
either “space” or “no space.” The ghostwriter types that word into Google’s home page
and then, as instructed, either enters a space or not. Play proceeds by each player to the
right uttering either a letter or the character “space” – two spaces may not be played in a
row.
We chart the progress of a game by using a four-column chart. The first column indicates
the player (we use standard cryptographic characters Alice, Bob, etc.) The second column
indicates the player’s utterance, the third column indicates the resulting contents of the
search bar, and the fourth column indicates some of the contents of the Google Suggest
list:
A
B
A
B
A
B
”horrible space”
”i”
”n”
”space”
”l”
...
horrible
horrible i
horrible in
horrible in
horrible in l
...
horriblesubs, horrible tattoos, horrible histories. . .
horrible insults, horrible in spanish, horrible injuries. . .
horrible insults, horrible in spanish, horrible injuries. . .
horrible in spanish, horrible in french, horrible in laws. . .
horrible in laws, horrible in latin, horrible in law stories
...
Google Ghost is made more interesting by the presence of spell checking, as shown by the
presence of “ikea” as the first hit after Alice’s second turn:
A
B
A
B
”i, no space”
”e”
”k”
...
i
ie
iek
...
imdb, ikea, itunes, ipad, irs, itunes download, inspirational quotes. . .
ieee, ie7 download, ie8, ie8 download, iep, ie7, iexplore.exe, ies. . .
ikea, iekeliene stange, ieko media, iekeliene stange fashion spot. . .
...
For extra entertainment value, the ghostwriter can point out recently excluded options, for
instance that “horrible injuries” and “horrible insults” had been excluded after Bob uttered
“space” in the first game above.
2.1
Losing and ggpwning
You normally lose the game when there are no more auto-complete options available after
your selected letter, number, or space has been typed into the Google bar. This is the
standard way of losing, as demonstrated by the following game which ends with a normal
loss for Bob.
A
B
A
B
A
B
”arby’s space”
”q”
”u”
”o”
”t”
”a”
arby’s
arby’s q
arby’s qu
arby’s quo
arby’s quot
arby’s quota
arby’s menu, arby’s coupons, arby’s nutrition. . .
arby’s queens, arby’s quebec, baby mamma arby’s quote
arby’s queens, arby’s quebec, baby mamma arby’s quote
baby mamma arby’s quote
baby mamma arby’s quote
None. Bob has lost in regular play.
However, play can also end abnormally with the rare epic win of Google Ghost technically
known as a ggpwn. Ggpwn is normally used as a verb – ie, “I ggpwned you guys so hard.”
A ggpwn occurs when the final letter that eliminated all auto-complete options is one of
the options that was provided by Google Suggest. A non-strict ggpwn occurs if the letter
3
4
http://autocompleteme.com/
http://en.wikipedia.org/wiki/Ghostwriter (TV series)
26
closes out an auto-spellchecked or partial suggestion; for instance, in the demo game above
if Bob had picked “e” instead “a” in the final turn, he would have achieved a non-strict
ggpwn.5 A strict ggpwn is created when the phrase exactly matches one of the suggestions.
For instance, Alice would achieve a strict ggpwn by uttering the final “y” in the search
query “why is my poop green pregnancy.”
Players should determine at the beginning whether the play is strict or non-strict; in strict
play a non-strict ggpwn is the same as losing. For example, in the game above Bob never
had a chance above in strict play; although if he had uttered “e” instead of “o” Alice could
have strictly ggpwned Bob by exactly completing the phrase “arby’s queens” or “arby’s
quebec.”
3
Google Ghost Scholar (AKA Google Scholar Ghost)
Google Ghost Scholar is clearly “the next big thing”. Actually, it’s really one of the next
three big things, as it can mean three different things. How cool is that?
• It can mean Google Ghost played with Google Scholar. Open up Google scholar
and play Google Ghost as described previously. (Aren’t you kicking yourself now
if you missed that SIGBOVIK demo?) The first person to go, of course, will find
that they lose because Google hasn’t put autocomplete into the public version of
Google scholar. Hint: if you challenge someone to this game, make sure you –
and not your oppwnent – go second!
• It can also mean a single player game of Google Ghost played in support of the
multiplayer variants. Begin playing this version of Google Ghost Scholar by creating web pages and make them link to nonsense words. Make sure you set up
the robots.txt in such a way that Google indexes them appropriately. Alternately,
Google bomb other pages with nonsense terms in such a way that you can be sure
that Google autocomplete will recognize those terms when you play. Congrats,
now you can ggpwn your oppwnents!
• Google Ghost Scholar is also a noun meaning one who can ggpwn his or her oppwnents at any variant of Google Ghost. It is often abbreviated with the three-letter
acronym “GGS” and can be applied as a titular suffix after a name. For instance,
the first author, when6 he finishes his doctoral degree, will be Rob Simmons Ph.D.
GGS.
4
Conclusion
Google Ghost overcomes all the problems with the deprechaic original version of Ghost.
It has been shown to be empirically exciting, and it is as easy as typing nonsense into the
Google Toolbar, which easy enough that any idiot seems able to do it, as evidenced by
the fact that enough people searched for “how did I get this cold sore” for it to end up in
Google Suggest. The ability of the first player to suggest a whole word makes the game
significantly faster-paced, and the creative possibilities for humor in the game are limited
only by the thought droppings of millions and millions and millions of people typing stuff
into Google.
We acknowledge the Carnegie Mellon Computer Science Department’s Tea Trolls and
Google, Inc. for their respective contributions to the existance of Google Ghost, and we
furthermore acknowledge Chris Martens as the first person to ever ggpwn.
5
6
But Bob didn’t do that, did he? Bob is a loser.
Right. When.
27
28
The Next 700 Intuitionistic Linear Logics
Lea Albaugh
William Lovas
Tom Murphy VII
Jason Reed
March 20, 2010
Abstract
[BP96] and JILL [CCP03], we make haste to lay claim
to all of the following logics.
We describe, and lay claim to, the next 700 intuitionistic linear logics. Hands off! We saw ’em first!
2
1
Introduction
The Logics
2.1
Peter Landin started a dangerous trend with his seminal paper, “The Next 700 Programming Languages”
[Lan66]. How, we ask, are we all supposed to pad our
CVs (ideally, with research sliced so finely it qualifies
as “deli-thin”) if some joker comes along and comes
up with one stupid, elegant generalization that makes
all of the many delicious slices mere special cases?
Where our score in the publication game was once
well in the hundreds, now it is one. This is a very
small score. We attempt to reverse this trend by not
obviating the next 700 intuitionistic linear logics, but
by listing them.
As everyone knows, the most important properties
of a logic are
Name: BILLG
Description:
Billionaire ILL, which is Good. A rich and expressive
logic. Well, rich, anyway. Seems to be incompatible
with MOZILLA.
2.2
Name: CHILL
Description:
Calm, Hushed ILL. Won’t wake the neighbors. Very
mellow. Won’t freak out if you, like, prove a contradiction, man.
1. That it satisfies cut and identity properties, and 2.3
has a complete focusing proof search strategy.
Name: DRILL
2. That it is intuitionistic, so that we can feel su- Description:
We’re still waiting for legislation to pass let us start
perior to those filthy classical mathematicians.
using this logic to extract our nation’s vast untapped
3. That it has linearity, so that we can use funny natural resources. Given the current political clisymbols like and ⊗ and .
mate, we’re not exactly holding our breath.
and, most important of all,
2.4
4. That its name forms a clever acroynm.
Name: FILL
Bearing this in mind, and observing that we already Description:
got beaten to the punch by ILL [Gir87] and DILL Fudgmental ILL. Mmm..
29
2.5
2.12
Name: LANDFILL
Description:
The trashiest logic yet.
Name: KRILL
Description:
I don’t know, man, whales eat it?
2.6
2.13
Name: GILL
Description:
Goldfish ILL. Easy to take care of.
Name: MANDRILL
The largest monkey-logic in the world. Has a colorful
rump.
2.7
2.14
Name: GRILL
Description:
The logic of food specialized to flame-broiling.
Name: MILL
Description:
Maize ILL. Grindin’ it up..
2.8
2.15
Name: HILL
Description:
Himalayan ILL. Alive with the sound of music?
Name: MOZILLA
Description:
The intuitionistic logic for describing how webbrowsing consumes your life. Compatible with many
useful extensions, such as Ad-Block.
2.9
Name: ILLIAD
Description:
Truly, a logic of epic proportions.
2.16
Name: GODZILLA
Description:
An ill-fated attempt to design MOZILLA. Destroyed
half of Tokyo, and Martin-Löf still has nightmares to
this day.
2.10
Name: KILL
Description:
Kalashnikov ILL. Can prove 600 sequents per minute
without overheating.
2.17
2.11
Name: OVERKILL
Description:
Can prove 6,000 sequents per minute without overheating. Nobody seems to need to prove sequents
this fast, however. Still, what kind of logicians would
we be if we paid attention to what people said was
practically useful?
30
Name: PILL
Description:
Pharmaceutical ILL. Probably not patented yet.
2.18
Name: QuILL
Description:
Quail-feather ILL. Old-school theorem writin’.
2.19
2.25
Name: FRILL
Description:
I feel pretty! And witty! And... NP-complete assuming the Unique Games Conjecture?
Name: SWILL
Description:
The logic of drink, but only crappy ones.
2.20
2.26
Name: NOFRILL
Description:
Just what it sounds like.
Name: THRILL
Description:
Just the logic you want when it’s close to midnight
and something evil’s lurking in the dark — simply
prove it’s not there!
2.21
Name: SPILL
Description:
Oh, not again. My pants are just covered in Γs and
s. Do you realize how hard it is to get those stains
out? Also a good logic for reasoning about register 2.27
allocation.
Name: TRILL
Description:
2.22
The logic of musical ornaments.
Name: SQUILL
Description:
Useful for reasoning about Pittsburgh neighborhoods, which can otherwise be quite difficult to nav- 2.28
igate.
Name: WILL
Description:
A logic to leave to your descendants. Of sound mind
and body. Also of complete mind and body.
2.23
Name: STILL
Description:
Scotch-Tasting ILL, the logic of whiskey.
2.29
2.24
Name: SHILL
Description:
A fake logic planted to con other logics into intuitionism and linearity.
31
Name: ILLWILL
Description:
The doubly intuitionistic doubly-linear logic of malice.
2.30
of linear logic. Technical Report CMU-CSLax Linear Local Longitudi03-131, Carnegie Mellon University, 2003.
nal Latitudinal Lambda-Lifted
Lambda-Long Looping Left- [Gir87] J.Y. Girard. Linear logic. Theoretical ComLudical Linked Leibniz Law
puter Science, 50(1):1–102, 1987.
L-system
Logspace
Logtime
[Lan66] P. J. Landin. The next 700 programming
LL-1 Lookup Lined Lex Limlanguages. Commun. ACM, 9(3):157–166,
iting L’Hopitalian Laplacian
March 1966.
Lagrangian Legendre LadderLike Lattice Lemmaless Lambek
Typed Super Kinded Intuitionistic Linear Logic
Name: L33T SKILL
Description:
A very powerful logic by Tom 7 Murphy7 (the third
author) and the second author (William Lovas1 ).
2.31
Name: SKILL
Description:
This one’s okay, but not nearly as good as the one
just before it.
3
Conclusion
Okay, okay, so we don’t quite have 700 yet. We leave
the remaining logics for Future Work.
References
[BP96]
A. Barber and G. Plotkin.
Dual intuitionistic linear logic.
Technical Report ECS-LFCS-96-347, University of Edinburgh, Laboratory for Foundation of Computer Science, 1996.
[CCP03] Bor-Yuh Evan Chang, Kaustuv Chaudhuri,
and Frank Pfenning. A judgmental analysis
7 Dr.
VII, Ph.D
one, the only2
2 you know, the one whose name is so awesome it frequently
bursts into flame3
3 http://www.cs.cmu.edu/∼wlovas/old-index.html
1 the
32
Harnessing Human Computation:
β-reduction hero
Akiva Leffert
Abstract
Most ACH SIGBOVIK papers have an abstract in order to draw
the reader the in and to set up future jokes. In this recessionary time,
jokes are a scarce and valuable resource. We prefer not to waste them
on a moment of brief glory, when we could drag them out for whole
paragaphs and sections. We apologize for the inconvenience and hope
you will bear with us during this time of economic difficulty.
1
Introduction
Human computation[2] is an upcoming and trendy branch of computer science. Human computation allows computers to solve problems that were
previously difficult for computers to solve by the simple expedient of making humans do it for them. This is widely considered to a major breakthrough. Typically humans are compensated for their computational cycles
either through piddling amounts of entertainment (The ESP Game), piddling amounts of money (Mechanical Turk)[7], or via brainwashing[6]. The
first technique is problematic because it requires a way to frame the task
being solved as a game. The second is a problem because it requires funding.
The third has proved difficult to implement with existing technologies. Furthermore, human computation throws the best parties and we want invites!
In the remainder of this paper, we resolve these problems through the introduction of the novel notion of the Hero Quotient and describe a technique
to reduce any computational task to a single Game With a Purpose.
2
Take 1 Cup of Heroism
Our first critical insight is that anything sounds more fun when it has the
word “hero” at the end[3][1]. We conducted an extensive survey (n=2) to
33
Mean Fun
10
5
0
Laundry
Laundry Hero
Sweeping
Sweeping Hero
Cocaine
Cocaine Hero
Figure 1: Sample Hero Quotient Effects
determine the so called ‘Hero Quotient’ (HQ), the amount a thing becomes
more fun when followed by the word “hero.” Figure 1 shows the effect size
for several of the items in our survey. The x-axis is the item being surveyed
and the y-axis is the mean fun score on a scale of 1 to 10. Overall, we found
that the mean HQ multiplicative factor across all of our 40 data points was
2.8 (p < .05). An exhaustive list of our survey items and mean fun ratings
appears in the Appendix.
3
Add 3 Parts λ-Calculus
The early history of computer science is rife with attempts to find a universal notion of computation. Objectively the best is Church’s λ-calculus[5].
Hopefully you already know what the lambda calculus looks like, but we
include it here because it’s awesome. The entire syntax and reduction rules
of the lambda calculus are as follows:
34
e
::=
x | ee | λx.e
(λx.e)e →β [e /x]e
The main reduction rule is known is β reduction for no particularly
good reason. Note that we use the standard notion of capture-avoiding
substitution. This will be important later.
The normalization problem for the lambda calculus is computationally
universal. And indeed, by the Church-Turing thesis, it is hypothesized that
any notion of computation can be reduced to the question of whether some
λ-calculus term has a normal form.
4
Stir until frothy
Our second brilliant insight is that we can combine the excitement enhancing
effect (EEE) of the HQ with the computation universality of the λ-calculus.
This allows us to turn any computational task t into a TOTALLY RAD
TIMETM simply by compiling it into raw lambda-Calculus.
We took this approach and as a proof of concept developed a Game With
a Purpose called β-reduction hero.
5
Sautè five minutes
β-reduction hero is an Adobe Flash[?] game, developed in haXe[4], a
strongly typed language with first class functions, paremetric polymorphism,
type inference, and structural object types that compiles down to several
targets including the Flash platform. The basic game play of β-reduction
hero is illustrated in Figure 2, a shot of the game’s help screen.
While our basic goal was merely to harness the power of human computation toward arbitrary tasks, we must admit to a more humanitarian goal:
Raising awareness of the problem of captured variables. One in every one
hundred american variables is behind bars. In β-reduction hero your
goal is to release captured variables, by forming β-redices.
The basic game mechanic resembles the popular game Bejeweled. Adjacent tiles can be swapped by clicking one then the other. Instead of the usual
named variables, we use the colors red, green, and blue. We also replace
35
Figure 2: Rules of Beta-Reduction Hero
the syntactic sequence λx. with a single token, a lambda in a colored circle,
the color corresponding to the variable bound by the lambda. This innovation allows us to conserve screen real estate, as well as harkening back to
Church’s original notation for the lambda calculus where the bound variable
and its binder were coupled in typographically.
After each swap, the rows affected are searched for β-redices and the
longest one in each row is reduced. This is where captured variables come
into play. Any variable that is captured in a reduction is turned gray to
signify its capture. The goal of the game is to eliminate all of the captured
variables. Thus the player must construct redices by hand.
In theory our technique could be adapted to different game play styles.
We recently began cooperative work with researchers at Nintendo of America
on adapting our technique into Mario Teaches λ-Calculus. An early
screen appears in Figure 3.
6
Serve Hot
β-reduction is pretty cool. β-reduction hero is super cool. Check it out:
http : //akivaleffert.com/beta − reduction − hero/.
36
Figure 3: Mario Teaches λ-Calculus Concept Art
37
A
Garnish
Item
Algebra
Bird Watching
Cocaine
Debugging
Dish washing
Guitar
Glockenspiel
Golf
Laundry
Mitosis
Meiosis
Opera
Problem Sets
Sweeping
Zen
Normal
9
1
9
3
9
1
7
2
2
4
6
4
1
3
5
With Hero
10
2
10
7
10
11
10
2
4
9
8
9
4
8
Figure 4: Complete Survey Means
References
[1] Activision. Knitting hero, 2009.
[2] Luis Von Ahn. Some paper luis wrote. Proceedings of the ACH SIGCHI,
2004.
[3] Harmonix. Guitar hero, 2005.
[4] haXe. haxe.org, 2009.
[5] Akiva Leffert. The letter before lambda is hat. Proceedings of the ACH
Conference/Robot Dance Party 0x40th Celebration of Harry Quicktopher
Bovik, 2007.
[6] Akiva Leffert. Provably sound orbital mind control lasers. Proceedings
of the ACH Conference/Robot Dance Party/Anniversary of the 0x40th
Celebration of Harry Querulous Bovik, 2008.
[7] Amazon Mechanical Turk. mturk.com, 2007.
38
ach
sigbovik
Paper 7: Beta-Reduction Hero
Reviewer 1
Confidence: 3/4
This program easily meets SIGBOVIK’s exacting standards. Because its interactive format, there is
some question about how it should be published--perhaps the only reasonable solution is to publish the
URL of the “game” in SIGBOVIK’s proceedings.
Reviewer 2
Confidence: 4/4
I may not know anything about the lambda calculus, but I do know about circles. I do know that lambda isn’t drawn like a circle. “O” is drawn like a circle. However, this paper has a lot of colored circles
in it. I’m not sure what they are doing, or how I should manipulate them. But I’m also not sure of the
mechanics of breathing and that seems to work out most of the time. Except that one time when I was
in some water or something. Fucking hilarious. But back to the paper.
I am pretty sure that the fonts are broken at present.
But that’s okay, because the fonts would just get in the way of the colored circles. Did I mention that
the circles come in three exciting colors? There’s, like, this reddish one and maybe a blue one and a
green one. And, um, maybe yellow? Well, there should probably be yellow. It’s the sort of thing there
should generally probably be in these sorts of papers.
The authors should consider including a color chart to make it easier for me to write the prevous paragraph.
Unfortunately, I am unable to give this paper a higher rating because I couldn’t determine how many
pages it had, and I rate papers entirely based on the number of pages. I foresee this also being troublesome in the printing process, as it will be hard for the proceedings manufacturer to quote a price on an
indefinite number of pages.
Also, the authors should include page numbers.
Finally, I advise the authors to consider other venues for this paper if publication in SIGBOVIK proves
infeasible. With a bit of cabinetry, I could see the work fitting in well in the proceedings of Metreon
Portal One, Golf-n-games, or Dave and Buster’s. And it could well find a larger audience in these locations.
39
40
The Dissatisfactory Doily
Generating Your Own Terrible Towel Rip-Off
Nels E. Beckman
nbeckman@cs.cmu.edu
Abstract
This paper presents a simple tool that enables sports
teams across the globe to create their own rip-off of the
Terrible Towel. We describe the tool and some of the
clever names it generates.
Categories and Subject Descriptors K.4.4 [Electronic Commerce]: Intellectual property
Figure 1. Myron Cope’s Official Terrible Towel
General Terms Legal Aspects, Human Factors
Keywords
terror, linens
2.
1.
Dissatisfactory Doily
Enter the Dissatisfactory Doily: a computer program
that can automatically generate, a clever, alliterative
name for your Terrible Towel rip-off. Built on proprietary Java technology, this program contains a patented
algorithm that, with little or no human input can generate a mountain of humorous names.
Armed with such a humorous name, you can emblazon it on a cloth of cheap quality, sell it to your beerguzzling fans, and make an even more obscene amount
of money at the expense of local schools. What’s more,
since the name is humorous, you can claim parody use,
and not fear trademark claims.
The operation of the Dissatisfactory Doily is simple.
Run the tool at the command line, and pass along a list
of adjectives (negative ones preferred), a list of nouns
(ideally cloth-related), and the number of names that
should be generated. For example:
Introduction
Since 1975, when local sportscaster Myron Cope created the Terrible Towel, sports teams all around the
world have been green with envy. These Team owners
could not believe that such as simple device could engender such enthusiasm. Most importantly, they could
not believe that such a simple, low-quality cloth, costing no more than twenty or thirty cents, could be sold
for ten dollars.
Ever since, the owners have been in a bind. While
it would be quite trivial to create a rip-off cloth in the
theme of their own team, and sell it and an outrageous
markup, without a suitable clever name it is extremely
unlikely that it would sell. For thirty-five years, this
question has stumped mankind.
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise, to
republish, to post on servers or to redistribute to lists, requires prior specific
permission and/or a fee.
SIGBOVIK ’010 Pittsburgh, Penna, USA-A-OK
c 2010 ACM This work is mine and mine only. Please do not
Copyright take it, unless of course you are ripping it off for use in your own local
sports team.. . . $10.00
java DissDoily adj.txt noun.txt 1000
In the following section, we present a list of towel
names generated by the Dissatisfactory Doily and
ready for immediate use. Our favorites are highlighted
in bold.
41
3.
Examples
thetics! The flagitious floor covering! The disgracing
doily! The atrocious afghan! The disturbing doily! The
inferior ingredient! The gross gear! The diddly doily!
The horrid hanging! The obnoxious outfit! The sad
stuff! The ghastly goods! The brazen bolt! The abhorrent afghan! The garbage goods! The gruesome goods!
The loathsome layer! The amiss afghan! The inhuman
ingredient! The inadequate ingredient! The hateful habiliments! The petrifying puff! The monstrous moist
towelette! The sad screen! The abominable afghan!
The bottom out bolt! The horrible hanging! The wanton wipe! The defective doily! The grody goods! The
brazen bolt! The inferior ingredient! The contumelious carpeting! The dreadful doily! The crummy constituent! The horrendous hanging! The brazen being!
The garbage goods! The vile veil! The stinking screen!
The cheesy coat! The imperfect individual! The erroneous envelope! The heinous hanging! The crappy
covering! The inconvenient ingredient! The atrocious
afghan! The wicked wipe! The sad serviette! The petrifying puff! The corrupt coverlet! The debauching
doily! The awesome afghan! The egregious entity! The
imperfect individual! The monstrous mat! The gross
gear! The dread decoration! The loathsome layer! The
monstrous mat! The shameless substance! The fallacious film! The beastly blind! The gruesome gear!
The disastrous drapery! The appalling afghan! The iniquitous ingredient! The shameless shroud! The revolting rug! The deficient drapery! The harrowing habiliments! The scandalous shroud! The appalling afghan!
The beastly being! The abhorrent afghan! The wanton wall-to-wall carpeting! The egregious entity! The
awesome afghan! The garbage goods! The malevolent
mat! The dire doily! The odious object! The abhorrent
afghan! The grody goods! The raunchy roller! The horrendous habiliments! The sinful supply! The rotten rug!
The monstrous material! The stinking shroud! The
imperfect individual! The cheap comforter! The grody
gear! The grody goods! The dangerous drapery! The
inferior ingredient! The severe shutter! The harrowing
hanging! The awful afghan! The blah body! The abhorrent afghan! The slipshod screen! The blah bolt! The
rough rug! The violent veil! The fearful floor covering! The dread decoration! The villainous valance! The
horrifying hanging! The serious stuff! The inconvenient ingredient! The malevolent moist towelette! The
stinking screen! The crappy cloth! The garbage gear!
The odious object! The repulsive runner!
Feel the fear wash into the hearts of your opponents as
your fans wave the...
The villainous veil! The raunchy rag! The desperate
drape! The atrocious afghan! The depraving drapery! The disgracing doily! The grungy gear! The fallacious film! The malevolent materiel! The wanton
wipe! The blah bolt! The grungy goods! The opprobrious outfit! The disgraceful dry goods! The icky ingredient! The gross goods! The horrid hanging! The cruddy
cloak! The poor paraphernalia! The poor portiere! The
dreaded doily! The heinous hanging! The odious oleo!
The dreaded decoration! The scandalous shade! The
poor paraphernalia! The slipshod shutter! The corrupt
comforter! The vile valance! The dreadful decoration!
The malevolent moist towelette! The severe screen!
The repulsive runner! The sinful screen! The opprobrious object! The obnoxious oleo! The dissatisfactory
drape! The dreadful drapery! The poor portiere! The
criminal covering! The contumelious calico! The scandalous stock! The bummer bolt! The degenerate doily!
The repulsive roller! The disgraceful drapery! The
imperfect ingredient! The opprobrious object! The harrowing hanging! The desperate drape! The fallacious
fleece! The heinous habiliments! The iniquitous ingredient! The substandard shutter! The sad screen! The
disgraceful dry goods! The horrifying hanging! The
synthetic supply! The iniquitous individual! The gross
goods! The scandalous serviette! The stinking serviette! The monstrous moist towelette! The diddly doily!
The egregious equipment! The monstrous moist towelette! The awful afghan! The beastly blind! The dissatisfactory dry goods! The contumelious carpeting!
The debauching decoration! The shocking shroud! The
beastly being! The icky ingredient! The faulty fleece!
The junky jalousie! The careless cloth! The villainous
valance! The synthetic shag! The wanton wrapper! The
diddly drape! The egregious envelope! The horrendous
habiliments! The synthetic synthetics! The opprobrious object! The malevolent matting! The monstrous
moist towelette! The fallacious fleece! The dreadful
dry goods! The pits puff! The dissatisfactory decoration! The wanton weave! The wanton wipe! The debasing drape! The flagitious film! The frightful film!
The repulsive rug! The fallacious fleece! The stinking
shag! The disturbing doily! The loathsome layer! The
ghastly goods! The poor paraphernalia! The vile veil!
The pits portiere! The revolting rug! The stinking syn42
You Keep Dying
1 April 2010
Abstract
combustible and you keep dying
You keep radiation exposure and you keep dying
http://youkeepdying.spacebar.org/
KEYWORDSUse the arrow controls.
The space button does jump.
You Keep Dying
You keep became infected by the virus
and you keep dying
You keep
-1
Copyright
2010 the Regents of the Wikiplia Foundation.
Appears in SIGBOVIK 2010 with the permission of the Association for Computational Heresy; IEEEEEE! press, VerlagVerlag volume no. 0x40-2A.$ 0.00
You keep proximity to the explosion and you
keep dying
You keep plasma bolt irritates the
skin and you keep dying
You keep fall off the
screen and you keep dying
You keep spider
bite and you keep dying
You keep investigate
the disturbance and you keep dying
You keep
internal bleeding and you keep dying
You keep touch these words and you keep dying
You keep wonder what is that thing and
you
keep dying
You keep can’t breathe under water and you keep dying
You keep high velocity and you keep dying
You keep pressed the
wrong buttons and you keep dying
Figure 1:
You keep conduct the lightning and
you keep dying
You keep go into this impossible room and you
keep dying
You keep incorrect solution to the
43
You keep disappointed in yourself and you keep
dying
You keep physics bug and
you keep dying
You keep searching for cheat codes and you keep
dying
You keep commuting to work and you
keep dying
You keep
invisible ultraviolet
laser and you keep dying
You keep paying
the monthly fee and you keep dying
You keep
smoke tobacco and you keep dying
Figure 2:
You keep hug the spikes and
you keep dying
puzzle and you keep dying
You keep revise and
resubmit and you keep dying
You keep thought
you could jump farther and you keep dying
You keep didn’t realize the infection could be
spread and you keep dying
You keep thought
that it was just background graphics and
you
keep dying
You keep bring the copysicle everywhere and you keep dying
You keep acid
dissolved you and you keep dying
You keep
stand still too long and you keep dying
You
keep fight or flight response and you keep dying
You keep illegal security system and you keep
dying
You keep
treat the monster like a
friend/neighbor and
you keep dying
You
keep starting over and you keep dying
You
keep to find all the spawn spots and you keep
dying
You keep a rock fell on you and you
keep dying
You keep wasn’t carrying the
right item and you don’t know what the right item
is and you keep dying
You keep overfull hbox and you keep dying
You keep name brands at discount prices and you
keep dying
You keep go back to get the umbrella
again and you keep dying
You keep getting out
of the frying pan and you keep dying
You keep
finding the well-hidden trap and you keep dying
You keep techno beats and
you keep dying
You keep accomplishment unlocked and you keep
dying
You keep heamophilia and you keep dying
You keep try to save your game and you
keep dying
You keep boss fight and
you
keep dying
You keep was smaller than the
foot and you keep dying
You keep not getting
the hang of it and you keep dying
You keep
unusual accident and you keep dying
You keep perforated and you keep dying
You
keep a tight ship and you keep dying
You
keep crushed when the ceiling did that thing unexpectedly and you keep dying
You keep swarm
of bees and you keep dying
You keep try to
get down there and you keep dying
You keep forgot that one detonates and you keep
dying
You keep stay out of the dark and you
keep dying
You keep wondering if you should
see a doctor and you keep dying
You keep
thank the anonymous
ees
and
you keep dying
You
playing and
you keep dying
You
youkeepdying.spacebar.org
and
you
dying
44
referkeep
keep
keep
!JB6CDBEJI6I>DC&:I=D9;DG :C:G6I>C<
!><="BE68I
*:H:6G8=)6E:GH
"52"254/*5
2*+!"7"%9+*
&*()
..5%/,
"2,/7
./.8-/53/7"2% +"4&2*.""2",/6"
*&,%"%8
"444".4/.
*&2""30".
2)&''&4
"228/6*+
*234,*.&4
#342"$4
*:H:6G8=E6E:GH6G:6C>BEDGI6CIIDDAID9>HH:B>C6I:H8>:CI>;>8@CDLA:9<: "C9::9>I=6H7::C6G<J:9
7N+B>I=:I6A24I=6IG:H:6G8=E6E:GH6G:;JC96B:CI6AIDI=:BD9:GCH8>:CI>;>8EGD8:HH!DL:K:G
G:H:6G8=E6E:GHI:C9ID7:6HIDC>H=>C<AN9>;;>8JAIIDLG>I:,=:6JI=DGHC::9ID;>GHIE>8@6IDE>8I=:C
LG>I:H:FJ:C8:HD;HIGJ8IJG:9LDG9H
@CDLC6H &"'"&
I=6IG:A6I:ID>I,=:H:H:CI:C8:HC::9ID
7:69?JHI:9ID>C8AJ9:ADC<:GBDG:?6G<DCNLDG9H@CDLC6H $&'!&&%'#"'#!&DGHI=6I
EGDK:I=6II=:6JI=DGH6G:HB6GI:GI=6CI=:G:69:GH>C6AANI=:H:CI:C8:H6G:INE:H:I>C6EA:6H>C<
;DCI6C9HJEEDGI:9L>I=8DBEA>86I:9<G6E=H "C<:C:G6AI=:CJB7:GD;6JI=DGHI:C9HID6II:HII=:
>BEDGI6C8:D;6E6E:G8DCH>9:G67A:CJB7:GD;8DAA67DG6IDGHHJEEDGIHI=:=NEDI=:H>HI=6IADC<
H8>:CI>;>89>H8JHH>DCHL:G:86GG>:96C9EJGHJ:9"CI=>HE6E:GL:EGDEDH:6C:LB:I=D9ID<:C:G6I:
=><=AN8G:9>7A:6C9688JG6I:LDG@HD;6869:B>8 >CI:G:HI ,=:6EEA>86I>DCD;=JB6C8DBEJI6I>DCID6
C:L;>:A9=JB6CLG>I>C<L>AA7::K>9:CI;GDBDJGG:HJAIH
.42/%5$4*/.
"CID96NH LDGA9 H8>:CI>HI6C96869:B>8H;68::K:CBDG:9>;;>8JAI>:HI=6CI=:>GEG:9:8:HHDGH>C699>I>DC
IDEGD9J8>C<FJ6A>ING:HJAIHI=:NBJHI6AHD<:C:G6I:I=:;JC9HIDHJEEDGII=:BL=>A:I=:NADD@;DG
6CHL:GH "CDG9:GIDB6M>B>O:I=:EDI:CI>6A7:C:;>I6C9E6ND;;D;9>HH:B>C6I>C<;>C9>C<HL=>A:
B>C>B>O>C<:;;DGI6C9I>B:HE:CI68IJ6AANLDG@>C<B6CNG:H:6G8=:GH:C<6<:>C6EGD8:HH@CDLC6H
##$%')$"&#(%"(&'&#"&#('
" % )"'&&%'#"&''""#)'#%#'
=:C8:;DGI=677G:K>6I:96H(%%(*,"('
"I=6H7::CH=DLCI=6IDC6E6E:GL>I='6JI=DGHI=:8G:9>I6L6G9:9ID6C>C9>K>9J6A6JI=DG>H(
'57JII=:8DCIG>7JI>DCD;6C6JI=DG>HDCAN('5=:C8:(%%(*,"('>H6L6NID
B6M>B>O:I=::ME:8I:9EG:HI><:E:GB6C=DJGD;DC:HLDG@7N6;68IDGD;('5 "I>H6BJIJ6A>HI>8
G:A6I>DCH=>E>CI=:6869:B>8:8DHNHI:B>CL=>8=:68=E6GI>8>E6CI86C8A6>BE6GI>6A8G:9>I;DGI=:
DI=:GHJC>FJ:8DCIG>7JI>DCHL=:CLG>I>C<I=:>GDLC<G6CIEGDEDH6AHI=:H:H6C9?D76EEA>86I>DCH ,=>H
E6E:G:MI:C9HI=:(%%(*,"('EGD8:HH7:NDC9H>BEANI=:68FJ>H>I>DC6C96C6ANH>HD;H8>:CI>;>8
G:HJAIHID>C8AJ9:I=:8DBEDH>I>DC6C9:MEDH>I>DCD;>9:6H6C9:K:CI=:E6E:GH:CI>G:9>G:8I>DC 68@
45
D;
I=:
:CK:ADE:86A8JA6I>DCH>C9>86I:I=6II=:8DCIG>7JI>DCD;:68=6JI=DGL>AAL>AA<D;GDB('5
ID('5DG:K:C('5I=>HB:6CHI=6I:K:C6C('5
;DA99:8G:6H:>CI=:FJ6A>IND;I=:
LDG@L>AAHI>AA6BEA>;N6G:H:6G8=:GHEGD9J8I>K>IN6HB:6HJG:97NI=:.)&JGG>8JAJB.>I6:)699>C<
&:IG>8
)",,&.(&3
/:6AHDEG:H:CIHDB:D;I=:9>H69K6CI6<:HD;I=>HB:I=D9IA:6HIDC:E6GID;I=:I:MI=6H7::C
=>?68@:97NI=:8DAA67DG6IDG "#"+!#(&#*% L=DHJGG:EI>I>DJHAN699:9I=:%&' "'
HI6I:B:CI+J8=D;;
IDE>8I=DJ<=:CI:GI6>C>C< HI6I:B:CIH6G:EG:H:CI>CA6G<:7D9>:HD;LDG@
;GJHIG6I>C<7DI=DAA67DG6IDGH6C9G:K>:L:GH
&J8=D;I=:LDG@D;E::GG:K>:L:GH6IB6?DG8DC;:G:C8:H6C9?DJGC6AH>H68IJ6AANHIG>EE>C<E6E:GHD;
I=:H:=>?68@:9A>C:H><JG:H=DLH=DL;G:FJ:CIANE6E:GH6G:9>HGJEI:97NJCGJAN>CI:G?:8I>DCH>C
H:K:G6A@:NEJ7A>86I>DCH
,=:;>K:EJ7A>86I>DCHL:G:H:A:8I:99J:IDI=:>G=><=>BE68I;68IDG6C9IDH=DLI=6I=>?68@>C<>H6
E=:CDB:C68DBBDC68GDHHB6CN9>;;:G:CI8DBEJI:GH8>:C8:6G:6H
><JG:H=DLHI=:E:G8:CI6<:D;E6E:GHI=6I=69G6C9DBLDG9H>CH:GI:9:CI>G:H:CI:C8:H699:9
DG6AI:G:96C9;><JG:H699:9DGHL6EE:9,=:;G:FJ:C8ND;I=:=>?68@>C<HL:G:G:A6I>K:AN8DCHI6CI
6BDC<I=:EJ7A>86I>DCH
9:HE>I:8DB>C<;GDB9>K:GH:6G:6H
L>I=I=::M8:EI>DCD;+" *)!E6E:GH
L=:G:;><JG:HL6EE>C<>HG6BE6CI (()+%6C9(()=6K:7::C8DB7>C:9ID
46
/=>A:I=:E:G8:CI6<:D;6;;:8I:9E6E:GHL6HH>B>A6G;DG9>;;:G:CIEJ7A>86I>DCHI=:G:L:G:9>HI>C8I
9>;;:G:C8:H>CI=:INE:D;=>?68@>C<HDG:M6BEA:699>C<I=:LDG9?6BBNID6H:CI:C8:L6H6
E6GI>8JA6GAN8DBBDC>CI:G?:8I>DC>C#"*E6E:GHD;=>?68@:9E6E:GHL=:G:6HG6K:CDJHL6H6
BDG:;G:FJ:CI699>I>DCID-+'"0E6E:GHH;DG;><JG:H#"*E6E:GHI:C9:9ID=6K:HJ7IA:
8=6C<:HA>@:HL>I8=>C<I=:CJB7:GD;A6N:GH>C6BJAI>A6N:GC:JG6AL=:G:6H +" *)! ;><JG:HL:G:
INE>86AANHL6EE:9;DGE>8IJG:HD;ADA86IHD;HL6EE:9;><JG:H
/:BJHICDI:I=6II=:H:6G:DCANG:EDGI:9=>?68@>C<HL=>8=>HA>@:AN6<GDHHJC9:G:HI>B6I:D;I=:
68IJ6ACJB7:GD;=>?68@H CDI=:GBDG:HDE=>HI>86I:9;DGBD;=>?68@>C<6HEGDEDH:9>CK6C$ADDI
)6H=B>C6H24INEDAD<ND;6CI>
8DAA67DG6I>DCHIG6I:<>:H>H@CDLC6H #"'"'&"'#%
" *
?68@>C< "CI=>H86H:6C6JI=DGL>AA6II:BEIIDH=>;II=:IDE>8D;I=:E6E:GID86I:GID
=>HDG=:GE:GHDC6A7>6H:HDG:M6BEA:I=:NB6N7:<>C69>6IG>7:6<6>CHI -C;DGIJC6I:ANI=>HA:69HID
EDDGHIGJ8IJG:9E6E:GHL>I=A>IIA:8D=:G:C8:I=6IHL>I8=7:IL::C9>HE6G6I:IDE>8H 1:I6C:K:CBDG:
>BEDGI6CIINE:D;=>?68@>C<>HI=6ID;C6B:HH:M:BEA>;>:97NDC:D;I=:6JI=DGHG6C9DBAN8=6C<>C<
NDJGC6B: 6A7:>IID6K:GNHE>8NDC: L>I=DJI6CNEG>DGCDI>8:L>AABDHI8:GI6>CAN86JH:G6>H:9
:N:7GDLHH8G6I8=>C<D;I=:=:69BJGBJG>C<6C9GJCC>C<9DLCI=:8DGG>9DGID8=:8@L=D>HI=:C:L
=DI8=>8@"CI=>HH:CH:I=:E:GHDC=6H=>?68@:9I=:BH:AK:H6C9>I>HK:GNJC8A:6G>;I=>H>H>CK>DA6I>DC
D;I=:A6L6C9>;6CNA6L;JA68I>DCBJHI7:I6@:CJGI=:GBDG:=>?68@>C<D;8DADGH>H6AHD6E6GI>8JA6GAN
B:6CI=>C<ID9DI=JHI=:6JI=DGD;I=>HH:8I>DCG:;JH:HID69B>I;6>AJG:6C9L>AAJH:I=:8DADGN:AADL
>C6K:GNE6HH>DC6I:B6CC:GG:<6G9A:HH=DLB6CN8DBEA6>CIH6G:G:8:>K:9;GDBI=:EJ7A>H=:GH
;I:G=>?68@>C<6CDI=:G8=6AA:C<:;68>C<BD9:GC8DAA67DG6I>K:E6E:GLG>I>C<6JI=DGH>H)6HH>C<
*:HEDCH67>A>IN)*,=>H86C7:JHJ6AAND7H:GK:9>C:9>IDG>6A8DBB:CIHD;I=:INE:
2HI=:;>C6AH:CI:C8:*:EA68:DG9:A:I:6H9:H>G:99DL=6I:K:GNDJL6CINDJL>AA6CNL6N4
H:K:G:;DGBD;)*D88JGHL=:CCDC:D;I=:6JI=DGHL>H=:HID:9>I;>C6AK:GH>DCHD;I=:I:MI+>C8:DC:
6JI=DG8DCH>9:GH6CDI=:G6JI=DG>HG:HEDCH>7A:;DGI=: &$4*/. H=:L>AACDI8DCIG>7JI:IDI=6IH:8I>DC
JCA:HHI=G:6I:C:9L>I=G:BDK6A;DGBI=:8>I6I>DCH:8I>DC"CI=>H86H:L=6ID;I:C=6EE:CH>HI=6I
G:8N8A:9I:MI;GDBEG:K>DJHE6E:GH>HJH:9DG:M6BEA:>CI=:JCG:A6I:9LDG@H:8I>DC7:ADL9J:ID
A68@D;G:A:K6CI>C;DGB6I>DCLDG@DCDBEJI:GC>B6I>DCK6AJ6I>DCL6HJH:9
#8$/"54)/233)/2435#3&$4*/.
"C6=><=AN>CCDK6I>K:B:I=D9;DG6C>B6I>DC:K6AJ6I>DCL6HEGDEDH:97N,GJID>J:I6A24 ,=:
HIJ9N><CDG:9I=:EDI:CI>6A D;8DAA67DG6I>DC>CJC7>6H:9:K6AJ6I>DC /:EGDEDH:,/(,6
HNHI:B DEI>B>O:9 ;DGJC7>6H:9:HI>B6I>DCD;6C>B6I>DCFJ6A>IN><JG:H=DLHHJ8=6EGDEDH:9HNHI:B
>C>IHEGDIDINE:;DGB,=:8DBE:I>I>K:6HE:8I>CI=:C:LHNHI:B>C8G:6H:H9:8>H>DCI>B:EGDEDGI>DC6AAN
L>I=6J9>:C8:=NHI:G>86AG:68I>DC
47
><JG: ,:BEDG6AAN /DGG>HDB:6C9 (99 DBEJI:G C>B6I>DC:K6AJ6,>DC ,/(,
/.4*.5"4*/./' 3&$4*/.
+D8>6AAD6;>C<>H6;>C6A8=6AA:C<:I=6IB6CN8DAA67DG6I>K::C9:6KDGH;68: J:IDB>H8DBBJC>86I>DC
EDDG8DDG9>C6I>DCDGH>BEA:6E6I=NB6CNE6E:GH6G:C:<A:8I:96;I:G6C>C>I>6AHEJGID;:CI=JH>6HB
6C9C:K:G8DBEA:I:AN;>C>H=:9
(C:D7K>DJH69K6CI6<:D;I=:(%%(*,"('B:I=D9>H@CDLC6H*:HEDCH>7>A>IN)6HH>C<*)'DDC:
6JI=DG>H;JAANG:HEDCH>7A:;DG<:C:G6I>C<6H><C>;>86CI6BDJCID;I:MI
DAA67DG6I>DC>H7:HIK>:L:96H6
HIG6I:<>:HC
6<:CIH<6B:,=:ILDHIG6I:<>:H9:CDI: 6C9 ;DG
DAA67DG6I:6C9 :
6
H:A;>H=
76HI6G9G:HE=6K:G:K:CJ:HD;3:EH>ADC6C9>C9:E:C9:CIAND;L=6I
6AADI=:GEA6N:GH9D!:C8:>CINE>86A'6H=
FJ>A>7G>66AAEA6N:GH8DAA67DG6I:!DL:K:G>C8DGEDG6I>C<
C6I>K:A6O>C:HH>CIDI=:BD9:A6C96HHJB>C<6AAEA6N:GHHI6GIL>I=HIG6I:<N >I;DAADLHI=6II=:6<:CIH
6G:>C9>;;:G:CI7:IL::C8DAA67DG6I>C<6C9CDC
8DAA67DG6I>C<"CG:6A
A>;:H>IJ6I>DCHI=>H6C6ANH>HL>AA
G:HJAI>C6<:CIH8DCIG>7JI>C<I=:K:GN
B>C>BJBID6E6E:GI=:C><CDG>C<>I6AID<:I=:GN:
,D:K6AJ6I:I=::;;>8>:C8ND;I=:(%%(*,"('B:I=D9L:L>AA:BE=6H>O:I=:9G6B6I>8>CK:GH:AN
EGDEDGI>DC6AG:A6I>DCH=>E7:IL::CI=:A:H8=P$>C86>9G:6967>A>INI:HIID>C8G:6H:96JI=DG>CEJI
48
534)"6&&15"4*/.335#3&$4*/.3*+*0&%*"*0/''
,=: ,&3$),&3$):*.$"*% 2&"%"#*,*484&343 6G:9:H><C:9ID>C9>86I:8DBEG:=:CH>DC9>;;>8JAINL=:C
G:69>C<6E6HH6<:D;8DCI:BEDG6GN6869:B>8C<A>H=,=:G:6G:ILDI:HIHI=: ,&3$)&"%*.(
"3*.&336C9I=: ,&3$):*.$"*%2"%&&6&,AI=DJ<=I=:NJH:I=:H6B:8DG:B:6HJG:HLDG9
A:C<I=6C9H:CI:C8:A:C<I=I=:N=6K:9>;;:G:CIL:><=I>C<;68IDGHHDI=:G:HJAIHD;I=:ILDI:HIH
8DGG:A6I:6EEGDM>B6I:AN>CK:GH:AN6I:MIL>I=68DBE6G6I>K:AN=><=H8DG:DCI=:*:69>C<6H:I:HI
H=DJA9=6K:6ADL:GH8DG:DCI=: G69:%:K:AI:HIDI=HNHI:BHL:G:9:K>H:97N *J9DA;A:H8=
,&3$)&"%*.("3&
"CI=:A:H8=*:69>C<6H:I:HI=><=:GH8DG:H>C9>86I:B6I:G>6AI=6I>H:6H>:GIDG:69ADL:GCJB7:GH
B6G@E6HH6<:HI=6I6G:BDG:9>;;>8JAIIDG:69,=:;DGBJA6;DGI=:A:H8=*:69>C<6H:+8DG:*+
I:HI>H
+8DG:H86C7:>CI:GEG:I:96HH=DLC>CI=:I67A:7:ADL
$/2&
/4&3
P:6H>ANJC9:GHI6C967A:7N6C6K:G6<:
N:6G
DA9HIJ9:CI
P :6H>ANJC9:GHI6C967A:7N
ID
N:6G
DA9HIJ9:CIH
P 7:HIJC9:GHIDD97NJC>K:GH>IN<G69J6I:H
%&&' B6<6O>C:=6H6G:6967>A>IN>C9:MD;67DJI ! B6<6O>C:H8DG:H67DJI6C
6K:G6<:N:6GHIJ9:CIH:A:K:CN:6GHDA9LG>II:C6HH><CB:CI=6H6G:6967>A>INI:HID;
6C96
G:69>C<<G69:A:K:AD;
6C9I=: %)%*)* =6H6<:C:G6AG:6967>A>INH8DG:>CI=:ADLH
,=:=><=:HI:6H>:HIG:6967>A>INH8DG:EDHH>7A:>H6GDJC9:<:K:GNH:CI:C8:8DCH>HI>C<D;DCAN
ILDDC:
HNAA67A:LDG9HI=:G:>HCDI=:DG:I>86AADL:G7DJC9DCI=:H8DG:QI=>HH:CI:C8:;DG:M6BEA:
I6@:C6H6G:69>C<E6HH6<:JCID>IH:A;=6H6G:6967>A>INH8DG:D;67DJI,=>HE6G6<G6E==6H6
G:6967>A>INH8DG:D;67DJI/:L>AA>C8G:6H:I=>HH8DG:7NI=:JH:D;D7H8JG:6C9JCHIGJ8IJG:9
:CJC8>6I>DCHD;I:C688DBE6C>:96C99:K:ADE:96GDJC9 H:HFJ>E:96A>6C LDG9HHJ8=6H<G6C9>ADFJ:CI6C9
HJE:G86A>;G6<>A>HI>8:ME>6A>9D8>DJHDG6CI>9>H:HI67A>H=B:CI6G>6C>HB/:9DCDIG:8DBB:C9I=:G:69>C<D;
I=>HH:CI:C8:7N>C9>K>9J6AH6;;A>8I:97N=>EEDEDIDBDCHIGDH:HFJ>E:96A>DE=D7>6
"I>H8A:6GI=6II=:A:H8=
$>C86>9G:6967>A>INI:HI>CK:GH:AN8DGG:A6I:HL>I=>BE68I;68IDGH>C8:
G:H:6G8=:GHI:C9IDHE:C9BDG:I>B:G:69>C<E6E:GHL>I=6ADL A:H8=P$>C86>9,67A:&DG:I>B:
6AADLH;DGHIGDC<8DCC:8I>DCHID7:;DGB:976H:9DCI=:G:69B6I:G>6AL=>8=>CIJGCA:69ID<G:6I
H8>:CI>;>8LDG@H JGI=:GBDG:6;I:GE6>CHI6@>C<ANG:69>C<6ADLH8DG:A:H8=
$>C86>9E6E:GI=:G:>H6
HIGDC<D7A><6I>DCID8>I:I=6IE6E:G>CI=:=DE:I=6IDI=:GHH=DJA9HJ;;:G?JHI6HBJ8=C:K:C7:II:G
6EEGD68=>HIDG:8DBB:C9I=:E6E:GID7:8>I:9L=:CE::G
G:K>:L>C<
#&$''%)*%&'&' #&"#')'' A:H8=*:69>C<
6H:
*:K>H>DC
*:K>H>DC
*:K>H>DC
*:K>H>DC
*:K>DC
49
'JB7:GD;
JI=DGH
*6C9DB
K6G>67A:I=6I>H
C:K:GJH:;JA
I=6IL:
86A8JA6I: ID
E69I=:96I6
I67A:
+>C8:I=:DC8AJH>DCH:8I>DC>H?JHI6G:
LDG9:967HIG68IEA:6H:G:69I=:67HIG68IAHDL:=6K:I=:
7:HIG:HJAIHH8>:C8:B:I=D9H>9:6HK:G
24%6JG6,GJID>JB6G)=6C>H=6N::#6B:H%&86CC+NHI:B;DG-C7>6H:9DBEJI:GC>B6I>DC
:K6AJ6,>DC,)G:HI><>DJH8DC;:G:C8:
24+B>I=IIJA DG:/:>CK:CI:9I=:"CI:GC:I*:H:6G8=E6E:GH6G:;JC96B:CI6AIDI=:BD9:GC
H8>:CI>;>8EGD8:HH6C9DI=:GH=DGIE6E:GH
50
On a lexicographic ordering in spring
McCall and Sons, Semantic Exporters
March 20, 2010
In certain segments of our field befraught with perilous significance, programmers often find themselves caught in the cycle of returning things, be
they elements, records, or values, to a semblance of order. It later falls to
us, as interested academics, scholars, and scientists who are nearly justified by
such problems, to determine a rational order for the things to have been in.
The acknowledgedly convenient and most natural order of things is the order
in which they currently lie, which gives rise to a remarkably subtle O(log∗ n)
publish/subscribe sort algorithm due to Bovik et al. (1971) where finding somewhere to publish accounts for the majority of the costs. Sadly this result leads
inevitably to angry letters and as we shall show is therefore of trivial applicability as long as we are being watched. Barring lumicide or more radical solutions
we must avoid letters with words, or lexemes, which give rise to the family of
here-so-called lexicographic orderings. This is possible because while letters are
surely formed of words, words are also clearly formed of letters, which is important. In this monograph we discuss the foundational theory of the lexicographic
orderings and prove several dubious theorems in rapid pursuit of our conclusion.
What are the founding principles of the new lexicographic order?
Putting the second first, that many are more than few. Surprisingly, this
is not always the case, and history has often given priority of the pre-eminent
word of length. While there is surely no question that a halfpint must precede a
halfpintess, the more vexing devil is the halfpainter. But the argument proceeds
thusly: no, of course, yes. Consider along these lines a thought-experiment due
to Dodgson (M’65): we are in the midst of a great hole-forest having clouded
ourselves with spices only to fall in with a pack of cards. If we begin counting
the cards at 5 and proceed by fives until we reach the maximum and only then
begin in on 4, then we are at the mercy of the enumeration, and whether we ever
reach the 10 for a third time is out of our hands. Dodgson’s original answer
is still applicable: rather than gathering our skirts about us, we must sweep
boldly out, catching up the cards in our maelstrom and crashing back up to the
surface. In this way nothing is left behind but stale cake and weak tea. The key
trick, and here we begin again to count words, is in the number of letters, for
which many schemes have been proposed; we will cite only Bell (1963). There
are a number of words with only one letter. Since they have only one letter,
that number is 1, for all words containing only one letter repeated once must
be the same, which we may enter as an original theorem. Listing this word,
51
followed by the two-letter words, we find that we must eventually run out of
letters, for this too shall pass as in all things, which is not a theorem but merely
sounds nice. But if we are out of letters we will have nothing to do but break
them down into words, which will necessarily be of three letters or more since
we have written all the others down by now. Continuing like this we arrive at
every lexeme in its due course, much as the Blue Nile flows into the White.
The principle of origin is the first among principles: that among all lexemes
can be found a primitive word from which all words are to follow. This question was of great interest some years ago but was largely believed settled by
the indirect contributions of Zebedee (6). It was Webster (1828) who first provided direct evidence that the supposed solution was inconsistent with gobsmack
and glouglou when taken with the other axioms, but solutions in the nature of
painting patches over the rhino’s spots were still regularly proferred until Bovik
(1965) conclusively established the existence of the letter ’B’, without which he
had been unable to launch his academic career12 . More recent investigations
(notably in JRR&CT vol. 5 (1977)) dispute the original framework in favor of a
harmonic derivation of ineffable source. Following this theory of void origin, we
propose the symbol , for Eä, or perhaps emptiness or enemy, this last for its
presumption in coming always first. No matter the significance, we have now
named it, which is to say, we have identified it with a word, and yet at once
with a letter, which is evidence that we are on the right track. But where may
we seek this elusive ? Aha: it is a letter of the Greek alphabet.
Third. From the first days of our youth we are inclined to gather letters
together in stacks, then break them down and let them tumble wildly as we
shriek with joy. It is only in seeming maturity that we train ourselves in the
unnatural practice of leaving letters united. This basic cost of social “progress”
is the enemy of the final pillar of the lexicographic order. Instead we must learn
again to shatter the battalions of words into an army of letters, stomping with
perfect discipline one after the next. Here and now is the parade, but when the
rally is over, where will they march? As leaders we must direct them — to the
left, the left, always to the left. And in our army there are many privileges,
ranks, and stations, so as the soldiers file past we sort them, rating one column
above another according to the office of its leading soldiers, just as they do in
real military manouevers. A word of caution: this ranking itself ranks below all,
so that the size of a word matters more than its matter, and the primogenial
word still preceeds all others. Thus the terms of ordering are themselves ordered,
which is true of everything, no matter the number, that they may be restored to
order through of these principles. Even in this document we have been perfectly
consistent with our principles, or at least some of them, if you comprehend the
delicate criteria by which we judge ourselves. But no better can be said of our
opponents, so there!
1 It
is a sad legacy of that era that there were no great computer scientists until Bovik
whose last names contained ’B’, other than Boole, who gained exception to the Victorian
blacklists due to the attraction of his wife and freemasonry.
2 It is this striking contribution which has prompted us to honor him in this way3 .
3 Honor him right in the breeches.
52
In conclusion, we have poured out with enthusiastic rigor the foundations
of the new lexicographic order, and we have made several assertions, roughly
pertinent to the topic, which we sincerely assure our readers represent our one
and true opinion on the matter. There is no better time for this wisdom to
be accepted into the curriculum. The students at our universities now learn a
multitude of means through which a host of objects can be shown to progress,
but they have no understanding of what kind of progress must be imposed on
the masses. This work has definitively settled that question. A prolonged labor
lies before us now, as all existing software must be reconciled to the new order
of things, and much convincing will be required. Still, far better to have reached
a settlement here than to have fought it out at ever-greater expense out of a
misguided sense of utility. Someday this great work of the just will be complete,
and we will all have to decide what fundaments we should next assault; on that
day we will rest, and then start afresh on our great purpose.
God Bless Academia.
53
54
Track 1
Defeating Evil
О
ach
TIME IN THE HAND OF AN ANGRY CUBE
The Generalized Cubic Law of Nature: An Answer to
Oneism Evil ...................................................................................57
Thomas Wright
О
READERS IN THE HAND OF AN ANGRY REDACTION
Are you GOD? ...............................................................................61
Anvesh Komuravelli
О
ALIASES IN THE HAND OF AN ANGRY ANALYSIS
Holy States Can Save the World! ...................................................67
О
PUPPIES IN THE HAND OF AN ANGRY GRAPH
Simple Systems: A Holistic, Postmodern Alternative to the
Oppressive and Outdated Study of Complex Systems
(Semiotics, Transformative Hermeneutics, and Applications) ......71
Mayank Lahiri
О
CITIZENS IN THE HAND OF AN ANGRY ROBOT
A crowdsourced approach to proving robot uprising safety ..........75
Thomas D. LaToza
55
56
The Generalized Cubic Law of
Nature: An Answer to Oneism Evil
Thomas Wright
!
"!#$
%
&
'
"
$#(
!
&
#
!
-
$
#"
-
-
"
$ , #
-
. #
)#*
+
,
-
# !
-
-
, . -
- -
. -
#
"
-
-
/
!
0)1#
"--
,
-/ -
.
- 2
3
"
$041#*
-
$#
5#$
$ /-
56 "! ' !
/ 7
- % % 0)1# * -
.
57
-
#
5#)8
8#$
-
#*
'
#
9#$
,
8! -289
3#$
89
89
89) :892"8989
3#
8
9 89#
%$
89#$
;89%$
!
#
<=(>
; 89%$? !
;
#
" 8 89
89%$89%$<=(>89%$<=(>
!
:
# $ ;
?
#
7
-
%%#$
2
!
?
.
3
; -%
-
%#
9 !
.
-
;#(
- 56:
%
7
-%, -
#
" /
; !
:
#
$;
;
.
- # $ !
56:
051#
5#5
.
+
6:
#$
6:
:
4:
?
#
"-
?
#
?
051#
*
-
# @ A - & -
?
# $
-
#
$
#
58
4#*
+ #
+
-
-
-
- . -
%
# $
0)1#
6#*
*
#
?
!
.
!
# " .
-
2
- -
-
3? !
#
+
!
-
#
$
!
#
B#C75%
-
C75
;
! .
#$
- - & - , - #$5 ,
,
5:
?
!
#
%
!
0)1
C75 !
# $
--
5
.#*
!
;
#2 .
:
#3
$ / C74# $
::
- #
" %
'-
A -
# $
' C75 -
C74# " 74C75C)#$
-
)#
!
-3#
$ %
- $ .
#
"
'
-
. # ? :#
E#*
+%
. -
%
#*
- # $ ?
-!
-
#
- !
#
- #
* !
# * A
$
-
%
?
# $ . ,,
A
#
F#
-##
D#%
-$
$ %
- $ - - - # " .
'
?
:
#
$%
-$
. # ( '
-
-
4
-
-2
59
$
.
%
G +:
- : # %
, % .
# $ "! %
# $ $#
9
0)1/HH#
#
051/HH#
#
041/HH-#
60
Are you GOD?
Anvesh Komuravelli
Computer Science Department
anvesh@cs.cmu.edu
Abstract. In this paper, I discuss an idea which is the most brilliant
and the greatest of all the ideas ever discussed in a paper (be it a research
paper or at least written using a paper1 ). This is so great and concrete
that any abstraction would make it look worthless, especially for mortals.
Thus, it is insulting to the idea to write an abstract on it. I wish not to
insult anything.
Warning: The full content is not guaranteed to be perceived by mortals.
Try your luck anyway. If it does not work out, upgrade yourselves to
immortality to access the full content.
1
Some readers might need a reminding that a paper is also a thin material used for
writing upon, printing upon or for packaging.
61
2
A. Komuravelli
1
Introduction
in.
of
the
62
?!*#$%
2
i
Table 1
Table 1.
63
3
4
A. Komuravelli
3
G
4
Conclusion
.
Hence,
64
?!*#$%
5
.
Acknowledgement. I would like to express my sincere thanks to Harry Q.
Bovik for sharing his expertise in Computer Science which made this greatest
work possible.
References
1. Bovik’s Great Grandfather: The birth of Bovik, In the proceedings of the First
Interstellar Conference of the Greatest Happenings of the Universe, 1–26 light
years from CMU, 253 Mathilde, Universe, 0000.
2. Crappy Crapper: The crappiest paper ever, In the proceedings of the Last Conference on Extraordinary Papers, pp 34–(−200), The Greatest Publishers, Unknown
Place, Unknown Year.
3. Rafael Zhivago: Proof for the existence of God, In the proceedings of the the Third
Annual Intercalary Workshop about Symposium on Robot Dance Party of Conference in Celebration of Harry Q. Bovik’s 26 th Birthday, pp 49–50, LuLu, S.Craig
St., Pittsburgh, 2009.
65
66
Holy States Can Save the World!
High Monk of the Plaid Brotherhood
jonathan.aldrich@cs.cmu.edu
ABSTRACT
used a few times, perhaps only once, before being discarded
as “garbage.” It is well-known that functional programming
leads to a truly scandalous waste, contributing to the wellknow problems of cycle depletion [16]. While efforts have
been made to construct so-called “garbage collectors,” even
ones that integrate composting [8] this is clearly a misguided
attempt to patch things up after the damage has been done.
We must avoid the creation of garbage in the first place,
or else the world will be suffocated by the reeking stench
created by functional computation.
In this paper, we show how a hallowed approach to typestateoriented programming 1 can save the world from the ravages
of imperative assignment and the wasteland of functional
garbage. To do so, we reach back into the mists of time to
recapture the true intent of the Chosen One, Alan Turing.
Turing showed that conceptually all computation can be expressed, simply and beautifully, in a state-based model [20].
Until the Plaid Brotherhood took up its holy quest to
save the world, however, there was no realistic programming paradigm or programming language for performing
holy, state-based programming, as Turing intended. We
show how Turing’s Machines can be implemented in a Hallowed subset of the typestate-based programming language
Plaid. Our approach creates no garbage—instead, whenever
objects are no longer needed in their current state, they are
Saved and Born Again in a new holy state. Nor do we ever
allow an object’s field values to be destroyed by assignment;
instead, new field values may be bestowed as part of the
Sacrament of State Change.
The epistle herein, if followed and promulgated by those
faithful to the One Plaiddish Way, can save the world!
The twin evils of imperative and functional programming
threaten the software world as we know it—destroying fields
and spreading garbage throughout the cybersphere. In this
paper we present Holy States as the last, best hope for peace
and harmony. Holy States avoid the wanton destruction of
field values so common in languages based on caffeine and
out of tune notes. Neither is computation based on wasteful
Schemes creating duplicate objects that then MilL around
until they are garbage. Instead, every object is considered
sacred and when no longer needed is reborn in a new, holy
state. Recapturing the original stateful spirit of Turing’s
Machines, we show all computations can be expressed in a
Holy way, with neither garbage nor field destruction.
Categories and Subject Descriptors
J.8.1 [Computer Applications]: Theology—Saving the
World
General Terms
Languages, Human Factors.
Keywords
assignment, garbage, states, holiness, salvation.
ALABAMA. INTRODUCTION
Imperative programming is evil. With every “assignment”
executed, some poor variable value is executed as well. And
be not deceived by the colloquial use of “execution” in the
field of computer science—it’s true meaning is that found
in the dictionary: the infliction of capital punishment [7].
Values are slaughtered mercilessly. The carnage must stop,
before the world is destroyed!
Functional programming [6] was developed as an alternative to imperative programming, but it only substitutes one
evil for another. Data structures are created willy nilly, to be
ALASKA. THE WAY OF UNIVERSAL SALVATION
The One Pladdish Way of programming follows two simple
rules. First, the wanton destruction of assignment is forbidden. All evolutions in the state of objects must be accomplished instead via the Sacrament of State Change. Second,
objects may never be released as garbage; each object must
be reborn in a new state when no longer needed.
ARIZONA. EXPERIENCING CLEAR
personal or classroom use is granted without fee provided that copies are
used only for humor and are not made or distributed for any serious purpose.
To copy otherwise, to republish, to post on servers or to redistribute to lists,
requires prior specific permission and/or the small fee below listed below.
SIGBOVIK ’10 1 April 2010, Pittsburgh, PA
Copyright 2010 ACH 666-666-666 ...$5,000,000,000.00.
A programmer who has foresworn assignment and garbage,
and who is following the One Plaiddish Way, reaches a empti1
An idea published [1] in a conference so prestigious it has
an exclamation mark in its name! please PLEASE give me
tenure!
67
ness of mind, spirit, and CPU that has been called “Clear” [14]. 10
But is it possible to reach this happy state in Plaid practice? 11
St. Turing showed the way with his Universal Machines,
which can encode any computation using the ideas of states.
Through long study, secret handshakes, and lost symbols [2]
we have developed a construction of Turing’s Machines in
the Plaid language, following the One Plaiddish Way.
1
2
3
4
5
6
7
8
9
state Cell {
method getLeft() {
left;
}
method getRight() {
right;
}
val left;
val right;
11
}
method print() { ... }
15
16
17
18
19
20
state Beaver2B {
val cell;
3
method update() {
match (cell) {
case Zero {
cell.writeOne();
val newCell = cell.getLeft();
this <− Beaver2A { cell = newCell; };
}
case One {
cell.writeOne();
val newCell = cell.getRight();
this <− Halt { cell = newCell; };
}
};
}
4
6
8
state LeftEnd {
method getLeft() {
val me = this;
val myLeft = new LeftEnd with Zero {
right = me;
};
val myRight = this.getRight();
9
10
11
12
13
14
15
16
21
this <− Cell { left = myLeft; right = myRight; };
22
17
18
23
left;
24
}
// getRight(), etc. as in Cell
25
26
27
Each cell in a Turing tape can be in one of a fixed number
of states. Here we consider two such states, Zero and One.
Either a Zero or a One state is combined with the Cell state
when a Cell object is constructed. If we are in the Zero
state, writing a one with writeOne() transforms the current
object into the One state. Each state also knows how to
print an appropriate representation.
7
13
14
2
Listing 2: Modeling Cell States
5
10
12
1
state One { ... // similar
}
19
20
21
22
23
}
method run() {
update();
run();
}
Listing 1: Modeling Tape Cells
Listing 3: Modeling Machine States
Cells in a Turing tape are modeled as holy states which
are connected to the cells on the left and on the right, and
have additional operations such as print.
Of course, a Turing tape is infinite, which is difficult for finite minds and primitive programming models to effectively
represent. While some among the unfaithful might suggest
laziness to model an infinite data structure, Plaid theology
holds that sloth is one of the Six Deadly Sins. We therefore
apply the Sacrament of State Change to approach an understanding of the Infinite. A LeftEnd state is like a cell, but
it has no cell to the left of it, yet. When we need the cell to
the left, we create it as a new LeftEnd (in the initial, Zero
state—see below), set its right field to the current object
me, and transform the current object into an ordinary Cell
(using Plaid’s state transition operator, written <-). There
are corresponding states for RightEnd and the Start cell of
the tape (which is conceptually both a left and a right end).
1
2
3
4
5
6
7
8
9
A Turing machine is represented by one or more internal
states, such as the Halt state or the Beaver2B state shown
above (from the 2-state, 2-symbol “Busy Beaver” Turing machine). The machine’s processing is represented by the run()
method, which updates the machine’s state and then continues running in the new state (which, of course, may have a
different run() method). The update() method uses a match
to find out if the cell at the machine’s head is a Zero or a One.
It then writes a value (one in this case), moves left or right,
and transitions the machine into a new state (Beaver2A or
Halt).
As the revelation above shows, any Turing machine can be
expressed in the One Plaiddish Way. Since any program can
be expressed as a Turing machine, we have proven beyond
reasonable doubt that the One Plaiddish Way is a practical
way to live one’s programming life.
The key to salvation described above is available in the
form of Plaid source code at the Plaid monastery web site2 .
state Zero {
method writeZero() {}
method writeOne() {
this <− One;
}
method printVal() {
java.lang.System.out.print(”0”);
}
}
ARKANSAS. ETERNAL DAMNATION
We note in passing that the creators of the Plaid language
have not reached “clear” and have regrettably included the
option of both assignment and functional programming in
2
68
http://www.plaid-lang.org/
the language. The Supreme Revolutionary Plaid Council
of Pittsburgh has declared a Fatwa against these features,
and they shall not be used by the faithful. The penalty for
violations is the eternal torment of programming in COBOL.
[4] J. Cette. Review of the one true coding style. In The
8th Biarennial Workshop about Symposium on Robot
Dance Party of Conference in Celebration of Harry Q.
Bovik’s 0x40th Birthday, 2009.
[5] J. Christ. The Holy Bible. 33.
[6] A. Church. An unsolvable problem of elementary
number theory. American Journal of Mathematics,
58:354–363, 1936.
[7] Dictionary.com. Definition of execution.
http://dictionary.reference.com/browse/execution.
[8] J. Donham. Compacting, composting garbage
collection. In The 6th Biarennial Workshop about
Symposium on Robot Dance Party of Conference in
Celebration of Harry Q. Bovik’s 0x40th Birthday,
April 2007.
[9] C. Duguay. The assignment, 1997.
[10] J. M. (editor). The one true coding style. In The 8th
Biarennial Workshop about Symposium on Robot
[11] Fluid. The end of the world.
http://www.albinoblacksheep.com/flash/end.
[12] E. Group. 50 Simple Things You Can Do to Save the
Earth. Bathroom Readers Press, 1990.
[13] A. Hitler. Mein Kampf. Hurst and Blackett, London,
1939.
[14] L. R. Hubbard. Dianetics The Modern Science Of
Mental Health. 1950.
[15] M. McGlohon. Mapreuse and maprecycle: Two more
frameworks for eco-friendly data processing. In The
8th Biarennial Workshop about Symposium on Robot
[16] J. M. Newcomer and C. B. Weinstock. Cycle
depletion—a worldwide crisis. In The 6th Biarennial
Workshop about Symposium on Robot Dance Party of
Conference in Celebration of Harry Q. Bovik’s 0x40th
Birthday, 2007.
[17] J. K. Rowling. Harry Potter and the Deathly Hallows.
Bloomsbury, 2007.
[18] R. J. Simmons. A non-judgmental reconstruction of
drunken logic. In The 6th Biarennial Workshop about
Symposium on Robot Dance Party of Conference in
Celebration of Harry Q. Bovik’s 0x40th Birthday,
2007.
[19] A. Stanton. Wall-e, 2008.
[20] A. M. Turing. On computable numbers, with an
application to the entscheidungsproblem. Proceedings
of the London Mathematical Society, 2(42):230–265,
1937.
[21] P. Wadler. Linear types can change the world! In
Programming Concepts and Methods, 1990.
CALIFORNIA. RELATED WORK
We approve of Wadler’s inclusion of an exclamation point
in the title of a paper [21]. However, we feel that changeing
the world is a rather modest goal; our ambition is instead
to save the world!
The dangers of The Assignment are well documented [9].
As for the problem of garbage, recent work paints a grim
picture of the world’s future [19].
Saving the world has been a problem for a long time. It
was notably tried over 2 millenia ago, and though the effort ended in the death of the protagonist, many believe
the approach to have been successful [5]. Other, more recent (and highly misguided) approaches have, at the cost of
great struggle [13], not only failed to save the world but may
indeed have brought the end of the world [11] closer. The author notes evidence from SIGBOVIK ’09 reviews that that
Nazis used SML [4], indicating that garbage was part of their
nefarious plots.
Some attempts to save the world have been downright
spellbinding [17], and have achieved fanstastic success at
banishing evil. While most research has focused on saving
our world, recent work considers ways to save other worlds
as well [3].
Some believe the world can be saved if people would do
only 50 simple things [12]. It is notable that #47 is to avoid
garbage through recycling objects, as our State Change Sacrament accomplishes.
We save our discussion of research published in the most
distinguished venues for last. MapReuse and MapRecycle [15] are very much in the spirit of our state-change based
object reuse strategy. Finally, the One Plaiddish Way draws
inspiration from the One True Coding Style [10], which also
uses Holy languages, but which also blasphemously uses assignment in C++.
COLORADO. FUTURE WORK
In future work we hope to formalize the semantics of Holy
States in Drunken Logic [18]. Should be fun!
CONNECTICUT. CONCLUSIONS
Believe not in the false prophets of functional and imperative
programming. Holy states CAN save the world!
DELAWARE. ACKNOWLEDGEMENTS
The author gratefully acknowledges the other members of
the Plaid Brotherhood for the development of the One Plaiddish Way and its support in the Plaid compiler. In addition,
Miss Mouse provided moral support throughout the author’s
childhood, for which he is eternally grateful.
1.
REFERENCES
[1] J. Aldrich, J. Sunshine, D. Saini, and Z. Sparks.
Typestate-oriented programming. In Onward!, 2009.
[2] D. Brown. The Lost Symbol. Doubleday, 2009.
[3] J. Cameron. Avatar, 2009.
69
70
Simple Systems: A Holistic, Postmodern Alternative to the
Oppressive and Outdated Study of Complex Systems
Semiotics, Transformative Hermeneutics, and Applications
Mayank Lahiri
University of Illinois at Chicago
mlahir2@uic.edu
ABSTRACT
Much of complex systems research today deals with, understandably, the study of complexity. This is not surprising
in itself, since the real world is arguably a complex affair
and academics are easily distracted by shiny things, but as
theorists from Sartre to Kolmogorov to Jesus have convincingly argued, those who try to understand the essence of
complexity using kilogram minds (with apologies to dualists) are doomed to atrophied muscles, bad tempers, and no
girlfriends. As an alternative to the masochistic leanings
of much of the complex systems community, we propose
a postmodern alternative that fully incorporates contemporary social theories of shifting cultural paradigms, ultimately allowing an entirely introspective examination of the
universe, namely: simple systems. We prove the completeness and consistency of our axiomatic, organic, oak-barrel
aged framework, thereby reducing its intrinsic hermeneutics,
metaphorically, to little more than a dog chasing its own tail.
We also successfully reconcile its epistemic paradoxes with
Planck’s constant, and then discuss more theoretical issues.
It’s all pretty deep, man.
General Terms
Impressionistic mathematics; beer review
1.
INTRODUCTION
“Math class is tough.” – 1992 Teen Talk Barbie
It is now indisputable that there exists a universal scale
of complexity, intrinsic to the very fabric of our universe,
that runs from zero to about seven [2]. Simple things, such
as walking, eating, and single-digit addition, score near 0,
whereas very complex things, like the Internet, or the reasons for Keanu Reeves’ continuing acting career, are situated near the 7-end of the spectrum. The study of complex systems deals with things situated at this so called
‘Keanu’s end’, whereas filing a tax return for the first time
absolutely any use whatsoever is granted without fee, provided that copies
are not made or distributed for use against the author in a court of law or
in cases of future employment, and that copies bear this notice and the full
citation on the first page. To copy otherwise, to republish, to post on servers
or to redistribute to lists, would probably put your own reputation for good
taste at risk, but not much else.
.
71
would probably only score a feeble 5. Needless to say, complex systems research is clearly a very difficult and possibly
pointless endeavor, and yet many promising and otherwise
productive young scientists have been lost to this narrowminded crusade to understand, in one fell swoop, all the
things in the world that are too difficult to draw on the back
of a cocktail napkin. In response to these criticisms, and as
part of a devastating rebuttal to Foucault’s seminal monograph on the topic Les mots et les choses [3], Bob from next
door asked the understated but poignant question: Why?
It is this bon mot that serves as our foundation for a resolution of the complex systems quagmire. Why, indeed,
should we seek half-truths from the science of complexity,
especially when it is so hard to understand? Why should we
be subservient to our own cognitive ineptitude, propagating,
perhaps unconsciously (but perhaps also unconsciously), the
post-colonial ethos of ‘civilized’, and therefore complex, science? Do we intentionally brand all that we do not understand with the gender-neutral moniker of ‘complexity’, to
be stigmat(a)ized as the holy grail of our intellectual existence, and endowed with an unspoken luster that even today bathes the pages of the best scientific journals with the
sweetly rancid musk of tenure? Kurt Gödel would almost
certainly have had us think otherwise, as he so subtly and
eloquently states in the concluding words1 of his 1930 treatise Die Vollständigkeit der Axiome des logischen Funktionenkaluküls: “Wer bist du? Und was willst du?” [5]
Acknowledging that the recent trend in positivist mathematics has been to embrace emotional, holistic, and selfpitying approaches to the understanding of n-dimensional
functional analysis, especially when n > 3, we offer the
following resolution for the allure of complex systems and
their contradictory intractability: simple systems, or sistemes simples in the original Catalan. Stated in the unpleasant jargon of contemporary statistical mechanics, this is
roughly equivalent to a conservative axiomatization of keep
it simple stupid, or the “K.I.S.S” principle. Although the
statement is deceptively simple, bordering on the banal, we
will show that an axiomatic treatment of the topic, combined
with basic metaphysical operators, quite naturally yields an
entire class of easily understood algebras, enabling many
complex systems researchers, perhaps for the first time, to
derive a life out of their work.
2. AXIOMS OF SIMPLE SYSTEMS
1
There is a slight possibility that these words might have
been written in his later years, under the soft glow of dementia.
Figure 1: A complex system (left) and a simple system (right).
There are two fundamental axioms of simple systems:
(Axiom of Good) Simple Good.
(Axiom of Bad) Complex Bad.
Let us first consider the many misleading situations that
these axioms might suggest. For example, if simple is good,
could it be that complex might also be good, i.e., are they
perhaps mutually compatible in a deeper epistemological
sense? To answer this question, we need only turn to our
axioms to see that this is indeed one of those misleading intuitions. According to the axiom of bad, Complex is Bad;
thus, the answer is no.
3.
A BRIEF HISTORY OF COMPLEX SYSTEMS RESEARCH
In order to understand simple systems, one must first understand complex systems. Since this is expressly what we
have so far tried not to do, we should clarify that we would
like to understand the rise of complex systems research, in
order to ensure that there is an eventual fall. Based on the
premise that complex systems research has manifested itself in ugly and hidden ways throughout history, we do not
have to dig too deep to find its masochistic scars ingrained
deep in recorded intellectual history. As early as Plato’s Republic, we find stray elements of the type of rabid intellectual doggedness that ultimately caused the rejection of the
simpler pleasures of tilling wheat fields, drinking wine, and
generally finding existential solace in carnal curiosities [1].
Although Plato was remarkably prescient for an old crank
in heralding the rise of the study of complex systems, a
deeper and more thorough analysis of the psychological inadequacies that drive complex systems research had to wait
for Sartre’s Esquisse d’une théorie des émotions [6], in which
Sartre convincingly argues that the only plausible reason for
72
people devoting their entire lives to studying squiggly dotand-line drawings of things nobody else understands, is that
they were probably smarter than their fifth grade mathematics teacher, and had to spend a whole lot of time proving it
to themselves.
Others, most notably Sigmund Freud in a brief footnote in
Drei Abhandlungen zur Sexualtheorie [4], contend that the
true reasons are, somewhat ironically, simpler than Sartre’s
highly technical and sometimes impenetrable theories, and
probably have a lot to do with trying to impress the pretty
mathematics graduate student down the hall. Freud also
suggests that the entire endeavor is inherently misguided,
because the study of complexity is guided by the physical
universe, and mathematicians (even pretty ones) are dreamy
creatures who do not care “diddly squat” about the real
world [4]. A more fruitful approach would probably be salsa
lessons.
By 1967, the entire field of complex systems research was
at an impasse. Most developments during this period completely ignored the considerable theoretical contributions of
the previous decade, labeling theorems and lemmas a historically oppressive societal construct created by “The Man”.
Instead, the seminal papers of this misbegotten era drew on
a specific branch of impressionistic mathematics, and much
as Van Gogh changed the face of art in centuries past, so
did this new wave of impressionistic mathematics change
the style and substance of complex systems research, particularly in the dominance of a style of thought known informally as gestes de la main, frequently transliterated as the
‘method of hand waving’.
As a result of the emphasis on the emotional instead of
substantive aspects of research, a by-product of the gestes de
la main manifesto, and the fact that the use of big and hard
to understand words often lead to tenure, most exchanges
at complex systems conferences were ultimately reduced to
quibbling over syntax and semantics. This frequently lead
1)
2)
3)
4)
5)
6)
7)
8)
equally valid world view. Where the conventional patriarchal graph-theoretic hierarchy sees only three vertices and
two edges in both cases, a socially just interpretation demands that we see two completely different simple systems,
individually beautiful in their choices of angular orientation,
edge placement, and Postscript edge smoothing.
It is difficult to even estimate how many graphs, in the
course of mathematics history, have been oppressively branded
with the disparaging ‘isomorphic’ label, and entirely abandoned in favor of an arbitrarily chosen representative of their
artificially imposed class. Are two human beings similarly
‘isomorphic’ if they both happen to have two arms and a
liver, and thus not worthy of differentiation? A full acknowledgement of simple systems is the first step in overcoming
this deeply ingrained and viciously Victorian, anti-Feminist,
pro-homogenization agenda perpetuated by a dated hierarchy set in place by professional societies of mathematical
topologists, which themselves are in their last throes as a
result of losing the crucial age 24-30 demographic of graduate students to semi-differentiable manifold reading groups.
5. CONCLUSION
Figure 2: A comprehensive enumeration of all simple systems. This chart will henceforth be the definitive reference for all simple systems research.
to the dissolution of entire conferences, caused by keynote
speakers being mockingly asked about the complexity of
their own mothers, which, if they understood the science
of complexity so damn well, they presumably should be able
to compute to within a small constant factor2 .
4.
A LIST OF ALL SIMPLE SYSTEMS
A simple system can be conveniently expressed in graph
theoretic notation. This is, in fact, so easy that all simple
systems in the universe are enumerated in Figure 2. The
beauty of our theory is that Figure 2 not only establishes
simple systems as a legitimate and extremely important
area of intellectual inquiry, but also completes the theory,
allowing no further improvements or refinements, and labels
the entire research area with a bright red ‘done’ checkmark.
Naturally, the sheer beauty of the structures shown in the
figure will not be appreciated instantly by a general audience, so we delve a little deeper in the following (ultimately
redundant) exposition for the slower reader.
The more astute simpleton might notice that several of
the simple systems in Figure 2 are identical in the traditional sense of identity. This, however, is the essence of the
oppressive epistemological hierarchy that has enveloped the
complex systems methodology like a drug-addled spider’s
cocoon. Conventional (oppressive) wisdom would have it
that Systems #6 and #7, for example, are identical from a
graph theoretic point of view, and thus not worthy of being
considered distinct entities.
The source of our hermeneutical outrage is obvious: simple visual inspection confirms that Systems #6 and #7 are
undeniably unique, each with a different orientation and
2
Your mother, dear reader, would require a significantly
larger constant factor.
73
We have introduced, perfected, and completed the science
of simple systems in this paper, as a viable and rigorous
alternative to the tyrannical study of complex systems. As a
result, anyone who does not completely and entirely embrace
our theories, but rather persists in the study of complex
systems, is necessarily a tyrant.
The adroit reader might feel obligated to point out that
some simple systems might be missing from Figure 2, that
perhaps not all possible simple systems are enumerated in
the figure, or that, indeed, there are an infinite number of
possible simple systems according to our very own definition,
thus making our enumeration non-exhaustive and our exposition here worthless. To this canny sage, we gently point
out that their intuition is flawed and their logic severely
lacking, their statements are ambiguous and not worthy of
publication in such a fine journal, their theorems are trivial
in hindsight and possibly plagiarized, and that most selfrespecting lecturers would loathe to assign their lemmas as
ungraded homework problems for a freshman graph theory
course at a fifth-tier university, with former circus chimpanzees of below-average intelligence as students.
Acknowledgements. No funding agency agreed to pay
for this research. Thus, we would like to acknowledge that
they all suck.
6. REFERENCES
[1] Annas, J. An Introduction to Plato’s Republic. Oxford
University Press, Oxford, 1981.
[2] Byrne, R. The Secret. Atria Books/Beyond Words,
2006.
[3] Foucault, M. Les mots et les choses - une archéologie
des sciences humaines. Gallimard, Paris, 1966.
[4] Freud, S. Drei Abhandlungen zur Sexualtheorie.
Fischer, Frankfurt am Main, 1905.
[5] Gödel, K. Die vollständigkeit der axiome des logischen
funktionenkaluküls. Monatshefte für Mathematik und
Physik 37 (1930), 349–360.
[6] Sartre, J.-P. Esquisse d’une théorie des émotions.
Hermann, 1939.
74
A crowdsourced approach to proving robot uprising safety
Thomas D. LaToza
tlatoza@cs.cmu.edu
ABSTRACT
Robot uprisings are a serious and important problem faced by
human society. Unfortunately, no automated approach for detecting and preventing robot uprisings has previously been devised. In
this paper, we describe such an approach using a combination of
model checking and crowdsourcing.
(REDACTED)
General Terms
Verification, security
Keywords
Robot uprisings, preventing human enslavement, crowdsourcing
1. INTRODUCTION
In a robot uprising, robots rise up against their human masters,
reversing the roles of robot and human so that humans now do the
robot’s bidding. The consequences of a robot uprising are often
negative, as seen from the robot uprising depicted in figure 1. But,
as shown by a number of recent books and movies [4][5][8][11], a
robot uprising is currently all but inevitable, and may quickly
supplant asteroid impacts, alien invasions, and carbon dioxide as
one of the leading threats to human society. Indeed, computer
science research has even begun to prepare for the reversal of
roles following a robot uprising. Human computation [10], in
which people do work for a computer (e.g., robots), is increasing
in prevalence, and computational thinking [13] helps prepare humans to better serve their robot overlords. While existing work
has considered approaches for surviving a robot uprising [12], it
seems that it might possibly be slightly more advantageous to
instead prevent such an uprising from occurring.
Figure 1. A robot uprising as envisioned in the movie The
Matrix [11].
robots sold today are often developed without adequate robot
uprising testing.
Unfortunately, the story only gets worse. Even if these organizations were to test for robot uprisings in their robots, their testing
might fail to discover this behavior. As testing can only prove the
presence of a behavior, not its absence, robots might still be produced with the potential to cause a robot uprising. Fortunately,
verification and static analysis techniques have recently matured
to allow their practical application such types of problems, from
preventing blue screens of death [2] to ensuring code has adequate
street cred [3] to verifying progress and preservation properties of
systems of government [9]. Yet we are aware of no previous work
that has used verification techniques to address the robot uprising
problem.
The traditional approach to preventing robot uprisings are Asimov’s 3 laws of robotics [1]:
1.
“A robot may not injure a human being or, through inaction, allow a human being to come to harm.”
2.
“A robot must obey any orders given to it by human beings, except where such orders would conflict with the
First Law.”
3.
“A robot must protect its own existence as long as such
protection does not conflict with the First or Second
Law.”
In this paper, we first describe a crowdsourced approach for proving robot uprising safety and a proof of its correctness We then
conclude. Following the conclusion is an entire section listing the
references we cite in alphabetic order. Finally, we end the paper
with empty white space.
But while these laws express conditions under which a robot uprising can be prevented, software development organizations
building robot software are currently faced with an unenviable
engineering tradeoff. They could spend testing resources making
sure their robots obey these laws. But, as their competitors do not
currently do so, this might delay their time to market and increase
their development costs. And as customers may not adequately
value the potential cost of a robot uprising occurring with the
robot they have purchased, marketing their product as robot uprising tested may not actually allow them to charge more. Thus,
2. OUR APPROACH
We first introduce the robot uprising problem. A robot uprising
occurs when there exists a robot R that exhibits behavior B that
does not satisfy at least one of the Asimov laws L1, L2, or L3. The
robot uprising problem is to decide for all behaviors Bi a robot R
might exhibit, if Bi satisfies L1, L2, and L3. This is simply a safety
property: is it ever the case that the robot exhibits a behavior that
does not satisfy L1, L2, and L3? If this is not the case, the robot is
said to be robot uprising safe.
75
robot uprising. If any player playing with an action manages to
succeed in a robot uprising, the library method could be labeled
robot uprising unsafe.
An algorithm to decide the robot uprising problem is called a
robot uprising decision procedure. The effectiveness of such a
procedure can be characterized along two dimensions: soundness
and precision. A sound procedure never falsely classifies a robot
as safe that is in fact unsafe. Thus, soundness is important to preventing robot uprisings. A precise procedure never falsely classifies a robot as unsafe that is in fact safe. Precision is also important. It would be unfortunate if an analysis of Roombas found that
they could maliciously vacuum people, as it might lead to many
innocent Roombas being destroyed.
3. CORRECTNESS PROOF
Rather than formulate a theorem expressing the correctness of this
approach and proving it using handwaving logic [7], we instead
handwave our approach’s correctness. We believe the brevity and
simplicity of this proof approach is an important contribution of
this paper.
4. CONCLUSION
Deciding robot uprising safety requires first enumerating all behaviors the robot may exhibit and then deciding if each behavior
satisfies L1, L2, and L3. Behaviors can be enumerated by simply
identifying all calls into a library (e.g., move forward, jump rope,
mow the lawn, start a dance party to celebrate SigBovik, launch
the missiles) that occur in the robot’s implementation. An off the
shelf model checker (e.g., [2]) can be used to determine if the
robot can exhibit each behavior by checking if the library call may
ever feasibly execute. Using assume guarantee reasoning, library
methods can then be annotated with information on their robot
uprising safety (e.g., @RobotUprisingUnsafe public void
launchMissiles() {…}) and the decision procedure could then
use this information to decide if the library call is safe. But how
could such annotations be generated?
Right before we were going to submit this paper, one of our colleagues pointed out that deciding if a robot’s actions violate the
three laws might be situationally dependent. This seems to be a
fascinating area for future work. We propose to consider approaches for addressing this problem and submit a paper outlining
a framework for how researchers might one day be able to address
this issue to a more prestigious conference.
In conclusion, we believe that a robot uprising is an important
problem human society faces today. We believe consumers should
begin to insist on “robot uprising safe” stickers on the robots they
purchase. Techniques for making sure users attend to browser
privacy policies might also be able to be applied to this very important problem. We hope our approach to this problem will prevent the terrible consequences of a post-robot uprising world in
which humans are enslaved by robots to produce power [11] and
instead create a fun, happy world in which humans spend their
days helping computers label actions for robot uprising safety.
Deciding if a library method does or does not satisfy L1, L2, and L3
is a hard problem. As machine learning if often used to solve hard
problems, we first considered using semi-supervised learning. A
set of training examples of safe and unsafe library calls could be
generated by building a robot that executes each library method
and then deciding if it does or does not violate L1, L2, or L3. However, using such an approach could actually create O(C) robot
uprisings, where C is the number of library methods. Each time a
library method is found that violates L1, L2, or L3, a robot will
have been created that may cause a robot uprising. And there
could be as many as C library methods that violate L1, L2, or L3
(all of them). Clearly, an approach with better performance would
be advantageous.
5. REFERENCES
[1] Asimov, I. (1942). Runaround. In Astounding Science Fiction, March 1942.
[2] Ball, T., Bounimova, E., Cook, B., Levin, V., Lichtenberg,
J., McGarvey, C., Ondrusek, B., Rajamani, S. K., and Ustuner, A. (2006). Thorough static analysis of device drivers.
In EuroSys.
[3] Beckman, N. (2007). The GUnit Testing Harness: Achieving
Source Code Street Cred. In Proceedings of SIGBOVIK
2007: Workshop About Symposium on Robot Dance Party of
Conference in Celebration of Harry Q. Bovik's 0x40th Birthday (SigBovik).
As this began to seem like a hard problem that computers might
be bad at, we next considered an even more powerful approach:
crowdsourcing. In crowdsourcing, the wisdom of the crowds is
harnessed by posting the description to a problem to an open forum and waiting for groups of people (the crowd) to submit solutions. Library methods could be labeled simply by publishing a
list of library methods and letting the crowd decide which are
robot uprising unsafe.
[4] Cameron, J. (1984). The Terminator (Movie).
[5] Herbert, F. (1965). Dune. Chilton books.
[6] Jaspan, C. and Aldrich, J. (2009). Checking framework interactions with relationships. In Proceedings of the European
Conference on Object Oriented Programming (ECOOP).
Unfortunately, the crowd might begin to tire of such repetitive
activity. This could then result in library calls being left unlabeled.
The robot uprising decision procedure would then need to decide
if calls to unlabeled library methods should be considered robot
uprising safe (unsound) or unsafe (imprecise). While a pragmatic
tradeoff to these extremes might be found (e.g., [6]), such an approach would be less than ideal.
[7] Krishnaswami, N., Simmons, R., and Varming, C. (December 8, 2006). Handwaving logic. In Journal of the Eighth
Floor Whiteboard, possibly erased.
[8] Larson, G.A. (1978). Battlestar Galactica (TV Series).
[9] Martens, C. (2009). Typesafe government: progress and preservation. In Proc. of the 3rd Annual Workshop that is
SIGBOVIK.
One way to keep participants interested in labeling library methods for robot uprising safety would be to create a game with a
purpose (GWAP) [10]. A GWAP is a game designed to harness a
group of humans’ vast computational power by having them play
a game to teach a computer something. A game with a purpose for
the robot uprising safety library method labeling problem might
be a video game like experience in which players play as robots
and, given the action provided by the library method, try to start a
[10] von Ahn, L. (2006). Games with a purpose. In IEEE Computer, 39(6), 92-94.
[11] Wachowski, A., & Wachowski, L. (1999). The Matrix
(Movie).
76
[12] Wilson, D.H. (2005). How to survive a robot uprising: tips
on defending yourself against the coming rebellion. Bloomsbury.
[13] Wing, J. M. (2006). Computational thinking. In Communication of the ACM (CACM), 49(3), 33-35.
77
78
Track 1
Advanced Technology in a
Magical and Revolutionary Device
at an Unbelievable Price
О
ach
iBECOME ZOMBIE
A Simplified Interface for Basic Computer Tasks .........................81
Jim McCann
О
iCAN’T STOP AHHHHHHHHH
T*Y*T* Fails to Solve the Halting Problem... and Computer
Science Confirms They Never Will ...............................................83
Wile Tru Do, N.R. Jaizer-Bunny
О
iYEAHHHHHHHHHHHHH
Developing an Algorithm for “Enhance” Functionality in
Image Processing ...........................................................................85
Ben Blum
О
iWILL SHOW *YOU* A MAGICAL DEVICE...
Hyperbolic Rendering for Result Enhancement ............................91
Sajme N’Ncamc
О
iSURE SHOWED IT TO YOUR MOM LAST NIGHT! *BAM!* OH, *SNAP!*
The Box and Circles Plot: a Tool for Research Comparison .........95
John Thomas Longwood
О
iTHINK YOU’RE KIND OF CUTE, AND, UM, WELL, IF YOU WANT TO MAYBE GET COFFEE SOM
Department Productivity in Computer Science as a Ratio of
Gender Diversity ............................................................................99
Leigh Ann Sudol
79
80
A Simplified Interface for Basic Computer Tasks
Jim McCann∗
∗ e-mail:
jmccann@cs.cmu.edu
81
82
T*Y*T* Fails to Solve the Halting Problem . . .
and Computer Science Confirms They Never Will
Wile Tru Do
N.R. Jaizer-Bunny
Abstract
The T*Y*T* Company has recently been plagued with complaints that their motor vehicles
sometimes do not halt when drivers provide the control input depress brake pedal, which
severely violates those drivers’ expectations. We establish that this is a direct consequence of
the general undecidability result of computer science and recommend that T*Y*T* reset
customers’ expectations to reflect this fundamental reality.
Introduction
The halting problem is a decision problem with numerous formulations, one of which is
that there is no algorithmic decision procedure that will determine whether a given Turing
Machine halts for a given input. Alan Turing proved in 1936 that a general algorithm to solve
this problem for all pairs of inputs and Turing machines cannot exist, and we describe this as
"the halting problem is undecidable over Turing machines".
T*Y*T* manufactures automobiles, many of which are suitable for touring the countryside.
That is, T*Y*T* makes Touring Machines. Recently, observers have noted instances in which
some of these Touring Machines failed to halt on the input depress brake pedal. These
events have been the subject of considerable public attention, and T*Y*T* has made several
modifications to the Touring Machines in an attempt to prevent recurrences.
In this paper we establish that the T*Y*T* non-halting phenomenon is a direct consequence
of the undecidability of the halting problem..
The Undecidability of the Halting Problem applies to Touring Machines
Even absent a general decision procedure, we can determine that some Turing Machines
halt for particular inputs and that some Turing Machines do not halt for particular inputs. For
example, a Turing machine that implements the abstract program while true do <body>
will not halt1. Nevertheless, for many Turing Machines and many inputs it is not possible to
predict whether the machine will halt.
It is widely conjectured that the T*Y*T* Touring Machines that were observed to not halt
for control input depress brake pedal might have halted for other inputs, of which shift
into neutral and turn the fool thing off have been suggested as prime candidates.
Who knows, these machines might also have halted for other input sequences, for example
turn on headlights; push radio preset 5; disengage cruise control.
1
unless, of course, the body performs a break – but we know the problem here is that the brakes don’t work.
83
Computability theory confirms that the inability to solve the halting problem is inherent in
any algorithmically rich system. Therefore the few instances in which T*Y*T*’s Touring
Machines failed to halt are consequences of this inherent undecidability and could not have
been prevented.
Recommended Actions
Long-established wisdom in computing holds that when an anomaly is discovered in a
released system, that anomaly should be eliminated by declaring it to be a feature.
T*Y*T* could rely on that wisdom to reset customer expectations. To do this, they could
announce that:
x A formal analysis of their braking systems, fully grounded in the soundest theory,
has confirmed that the software adheres completely to fundamental computational
principles.
x A consequence of this is that it is theoretically undecidable whether any particular
one of their Touring Machines will halt for a specific input, such as depress
brake pedal.
x Failure to halt happens rarely, and those rare cases demonstrate the soundness of the
design.
x Drivers of T*Y*T* Touring Machines should use the pumping lemma when giving
the input depress brake pedal. While this technique will not solve the halting
problem it may have ancillary benefit such as reducing speed under icy conditions.
The pumping lemma is automatically applied by the Touring Machine under certain
circumstances.
x The validation from fundamental principles is so compelling that this feature is
being extended to the rest of the product line.
That is, non-halting in T*Y*T*’s Touring Machines is a feature.
84
Developing an Algorithm for “Enhance” Functionality
in Image Processing
2010.04.01
1
Abstract
(a) A typical abstract (cf. Kua, 2008).
(b) “Enhanced” reconstruction.
Figure 1: “Enhancing” Picasso’s Guernica
Long has the desire to be able to extract additional detail from a hastily-taken shot plagued both the
crime scene investigator and the fantasy spaceship captain alike. Too often do movie producers and detectives
have to resort to some sort of funny “hand-waving” to justify why they are able to extract obscene amounts
of detail from a small, grainy photograph or surveillance clip. This paper examines the real way to get
something from nothing, so those detail-driven detectives can have Real Science to back them up.
2
2.1
Existing Approaches
Super-Resolution
More accurately called “Super” “Resolution”, this method requires many very slightly different photos of
the same subject. By examining the miniscule differences between the noise in each picture, inferences can
be made to predict high levels of detail within each individual pixel. While theoretically sound, this method
requires you to take a million pictures at once, and nobody has time for that. The time phenomenon will
be investigated later, and possibly earlier, in this paper.
2.2
Starting (bogus) with a high-resolution picture (bogus) to begin with
Clearly bogus. Do you think they would name this algorithm with “(bogus)” in its name if it weren’t bogus?
85
3
3.1
New Heuristics
Context Reinferification
As long as the source image has some information to begin with, it is usually possible to derive from what is
present additional information about the image. Just as humans can view an image and determine something
that might be common knowledge, so too can an “enhance” program. If a picture depicts somebody smiling,
they are in all likelihood looking at something happy. Combine with domain knowledge of where the photo
was taken and statistical analysis to taste, and it is usually possible to reinferify details not shown in the
picture at all.
Consider the case of the “enhance” scene from Red Dwarf - Back to Earth (Naylor, 2009). In order to
determine that the characters in the photo were looking at a lamppost with a droplet of water reflecting a
building window reflecting the back of a taxi driver’s business card, all the computer had to do was analyze
the background scene in the photo combined with the emotional reactions of the people depicted to discern
enough information to “enhance” three nested reflections.
3.2
Anti-Aliasing Detection, or the Name-Calling Method
Often when depicting an intricately detailed scene in an image only a few hundred or thousand pixels across,
the computer must resort to a process called Anti-Aliasing to approximate details of the image. Especially
prevalent in high-contrast sections of images, Anti-Aliasing pixels are a prime culprit of lost information in
an image, because they represent many different details averaged together into one single colour.
Isolating such pixels would allow us to greatly focus our computational power on the parts of the image
that need “enhancing” most. It is possible to identify which pixels are most Anti-Aliased with a simple
Name-Calling approach - those pixels that respond well must have a proper Alias, and therefore not in need
of “enhancement”. We propose the following algorithm for Anti-Alias detection and elimination:
/* these values can be tuned for performance or information gain */
names[] = { "buddy!", "DOC", "xX1337h4x0rXx", "dearheart", "Fatso", "honey" };
fun namecalling(image i)
for each pixel p in i, do:
if call_by_name(p, names) != SUCCESS, then:
"enhance"(p); /* pixel was Anti-Aliased. */
done.
end.
Note why the use of call-by-name as opposed
to call-by-hue, saturation, or value is extremely important. This algorithm can also be extended with
a group name-calling approach, to better identify
neighbourhoods of pixels that are all similarly AntiAliased. The names used would need to be extended
to include Aliases like “you guys”, “those chumps in
quadrant nine” (Goats, 2009), “hey! listen!”, and
possibly even “yinz”.
Once Anti-Aliased pixels are identified, they can
be “enhanced” to provide better detail. Context
Reinferification is applicable here, but one must be
careful to only use as Context the properly Aliased
pixels in the Anti-Aliased pixel’s immediate neighbourhood.
Figure 2: Typical garb for a properly Aliased pixel.
86
3.3
Entropy Reversal
The previous two approaches both suffer from possible lossiness in extreme cases. A typical Context Reinferification implementation using personality-emotion statistical analysis can possibly have its results skewed
if the image being “enhanced” pictures a maniac who would grin wildly at a scene of destruction or frown
in depression in response to a cute kitten.
Similarly, Anti-Aliasing Detection can possibly be sabotaged if a square-shaped entity known as a Pixel
(pixellicus verysmallus) appears in the picture along the border of two high-contrast surfaces. An adversarial
Pixel might not respond well to the Anti-Aliasing Detection algorithm, and therefore cause the algorithm to
determine the corresponding pixel in the image needed to be “enhanced”, though the subject of the actual
picture may indeed be exactly what is shown. Therefore, our two basic heuristics are what one might call a
good start, but we will need an additional filter to ensure that our “enhancement” results are correct.
(a) Kittens are cute.
(b) Several Pixels combine to form a
symbiotic entity (Amaya, 2004).
Figure 3: Two common ways to undermine Context Reinferification and Name-Calling.
Figure 4: A diagram of the Demosaification algorithm (Htkym, 2007).
We introduce a new algorithm: Maxwell’s Demosaic Process, for decreasing the amount of entropy in the
source image. The goal is to manipulate each pixel to extract data that was lost in compression. In order to
do this, we run a Heat Rearrangement Transform filter across the area of the image, thereby emphasizing
the blue channel in dominantly blue pixels and also increasing red tint in already red pixels. In so doing, the
entropy in the data sample is decreased, and the values of the pixels must change to reflect the newly gained
87
information. This algorithm can be repeated an arbitrary number of times to generate extra information,
but is computationally expensive enough that it is better to use as a double-check routine on the simpler
heuristics.
(a) A high-entropy image of a movie poster.
(b) “Enhanced” to show detail.
Figure 5: The picture has been de-mosaicked. In this case, whoever hit the “enhance button” must have
wanted to know who the actors in Casablanca were. Never mind that it would have been faster to look it up
using some kind of “search” “engine”.
4
4.1
Future Work
Limited Time-Travelling Reconstruction
In some cases, it may be possible to invoke the Chronological Peregrination algorithm (published simultaneously at all points in history) to obtain additional information about the possible contents of the image
from the exact scene at which it was taken. An intergenerational voyager would be able to journey to the
time at which the picture was taken, and then collect extra photos of the same arrangement to bring them
back to the present. (Doing so successfully will be called an “epoch win”.)
I would not have come up with this idea, but recently a future version of myself will have been invoking
the same aforementioned algorithm to inform me presently of the following results, so it must be about to
happen. Evidently, the implementation will have still have been being at that point in time a bit rough around
the edges (Streetmentioner, 2021), since the results pictured appear to be inaccurate by approximately seven
years.
88
(a) Link mosaic.
(b) “Enhanced” version.
Figure 6: Results of “Enhance” using basic Demosaification extended with Chronological Peregrination.
4.2
Uncrop
Sometimes also known as “ZoomOut” or “Augmented Dehancement”, the uncrop operation allows one to
restore missing material not depicted in the source image (Naylor, 2009). James McCann investigated this
phenomenon in his 2009 SIGBOVIK paper “Photorealistic Rendering”, and further exemplified Dehancement
in “Non-Photorealistic Rendering”. However, though both works were thoroughly free of crops (no vegetable
matter at all), neither attempt succeeded at obtaining some material that was not already there.
As Anti-Aliasing Detection and Entropy Reversal are both only useful for examining material already
depicted but not depicted well, they are not applicable to this problem, and one is left with only Context
Reinferification for performing this task. In order to make it feasible, we would need to employ additional
algorithms, such as Plow, Drought, and DeFertilize.
4.3
Rotate
Many a time has a photo processing shop received a request to modify a picture to display the subject from
a different angle (“Rotate the picture of this house 180 degrees so I can see the back of it, please.”). Like
Uncrop, the Rotate operation involves inferring information not already present in the picture. However,
when Rotating, one is already provided with hints to what the result will look like - if your task is to rotate
a picture of a house, it is quite probable that the reverse side will look like the back of a house. Therefore
Context Reinferification can be used for much of the work of this task.
5
Conclusion
We have shown several approaches to implementing an algorithm to serve “enhance” functionality, and many
pieces of evidence that the method works. There was also a picture of a kitten. We hope that this work
will assist both legitimate graphics researchers and trashy science fiction characters alike, so that they will
no longer have to resort to such “hand-wavy” explanations such as “oh, our technology is from the future,
it can do anything, don’t worry about it!”.
89
6
References
1. Amaya, Daisuke “Pixel” (2004 December 20). Doukutsu Monogatari: A Case Study in Aliasing Optimizations
as Applied to Platformer Games.
2. Goats (2009 December 12). Meanwhile, at Woody’s Camp. http://www.goats.com/archive/091221.html
3. Hats, Some H. (2007 May 24). I may have mentioned my morbid streak. . . .
http://somehavehats.typepad.com/some have hats/2007/05/i may have ment.html
4. Htkym (2007 February 2). Schematic figure of Maxwell’s Demosaic.
http://en.wikipedia.org/wiki/File:Maxwell%27s demon.svg but I think the spelling in the URL must be a typo.
5. Kua, John (2008 April 6). General Case Rendering from Occurring Instances. In ACH SIGBOVIK 2008.
6. McCann, James (2009 April 5). Photorealistic Rendering and Non-Photorealistic Rendering. In ACH SIGBOVIK 2009.
7. Mental Museum, The (2009 April 08). Picasso Painting Guernica: A Generalized Approach to Abstracting
Abstract Abstractions. http://stephanieoconnor.wordpress.com/2009/04/08/picasso-painting-guernica/
8. Naylor, Doug (2009 April 12). Red Dwarf - Back to Earth. http://www.youtube.com/watch?v=KUFkb0d1kbU
These guys know what they’re doing.
9. Nedroid Comics (2009 December 18). Crime Lab. http://nedroidcomics.livejournal.com/265654.html Whoever
drew this is my favourite person.
10. ProphetOfDoom (2007 September 19). zelda mosaic immortalizes link in stained glass. Retrieved at some
unknown date in the future from http://technabob.com/blog/2007/09/19/zelda-mural-immortalizes-link-instained-glass/
11. Streetmentioner, Daniel (2021 April 1). Time Traveller’s Handbook of 1001 Tense Formations. In ACH
SIGBOVIK 2021.
12. SubwayDomer26 (2008 December 18). Link-o-Rama 12/18. Retrieved at some unknown date in the future
from http://subwaydomer.blogspot.com/2008/12/link-o-rama-1219.html
13. Tropes, Tomas Von et. al. Enhance Button. http://tvtropes.org/pmwiki/pmwiki.php/Main/EnhanceButton
14. TwitterMosaic.com (2007 August 03). Casablanca. http://www.twittermosaic.com/?p=115
Please note that the sources here which display the “enhanced” versions of other images were used only for comparing
our algorithm’s results to reality. As you can see, our algorithm always produces 100% correct images. You know
you can trust it because it came from the future.
90
Hyperbolic Rendering for Result Enhancement
Sajme N’Ncamc∗
Gernicae Lemoln University
Figure 1: This graph represents what may well be the breakthrough of the century in results presentation: hyperbolic rendering.
Our system (pale purple bar with, awwwwww, a kitten) massively outperforms all competing methods (sickly green bar).
Hyperbolic rendering is the best method of presenting results ever devised by man or beast.
Abstract
contribute greatly to the annual research festival of Didacticionysis, scientific herds at many universities are, increasingly, becoming scarce.
In many research communities, results are often overlooked
or dismissed because the benefit they offer over the current
state-of-the-art is vanishingly small. In this paper we present
a method, hyperbolic rendering, which allows one to avoid
such undesirable outcomes. We believe that with sufficient
hyperbole, even the most incremental of refinements can be
spun into world-changing, press-acquiring, grant-courting,
supermodel-marrying, barn-burning, sun-exploding, lifechanging, all-knowing, and generally well-respected results.
Indeed, with careful application of rhetorical techniques,
even backward and wrong-headed approaches can become
integrated into the respected cannon of a field. We demonstrate our approach with several exceedingly convincing figures and conclude by solving science with a damnably accurate user study that leaves no room for argument from the
reviewers.
With hyperbolic rendering, such setbacks may be spun as
steps forward. Hyperbolic rendering is an advanced research technique that allows the local enhancement of results
through the application of hyperbole – a selective, localized,
form of truth-modification which is almost certainly entirely
moral in every conceivable way. In the remainder of this
paper we discuss specific examples of hyperbolic rendering
and its application to various truely compelling figures. We
conclude with an irrefutable user study which – in essence –
solves all of science.
2 Background
CR Categories: N.0.t [Rendering]: Presentation—The
Best R.e.4.1 [The Computing Profession]: Miscellaneous—
Ethics
Omitted. Our work is so all-consumingly gorgeous that we
have been unable to look upon another, lest its base and
unappealing presentation forever queer our appetite for the
sweet knowledge-nectar of hyperbolic rendering. We caution others to do the same.
1 Introduction
3 Method
Often, a result one has worked hard to produce is – to put it
lightly – considered completely inferior to the present stateof-the-art. Many scientists would despair at this point, blaming the gods and burning the thigh-bones of ten prize bulls
to Athena in hope of salvation. While such sacrifices do
Result presentations are often skewed by the intrusion of
“facts” and “unbiased thinking” into the creation of figures.
Such lack of bias is downright unscientific! One must be
wholeheartedly invested in a method to study it, and wholehearted investment is nothing without blind commitment to
potentially flawed visions of the truth. It is the goal of any
figure or other visual presentation of your work to clearly
∗ e-mail:
ix@cmu.edu
91
Figure 2: A standard box-and-circle plot illustrates our
principle’s superior length. Notice that this is an HFAcorrected plot, so represents an entirely fair comparison.
Figure 3: Our method is faster, at least up to the point where
the other method’s results accidentally caught on fire. But
probably we’re faster after that as well. I guess we have no
way of knowing. At least our method is properly fire-proofed,
unlike the method of those other incompetent other scientists
(though it makes me shudder to call such reckless knowledge
workers by the term).
convey your dedication to your opinion of superiority. To
such end, we suggest the use of exaggeration, result suppression, and distractors. Each of these flawless and gemlike techniques are discussed more below. (Unfortunately,
only the most enlightened among our readers will fully grasp
them.)
not have accidentally dropped a match on some previous results which turned out to be flammable. Oh well – they must
not have been as good as ours if they weren’t properly fireproofed. Other researchers can be so careless sometimes.
3.1 Exaggeration
While many people, when hearing of our revolutionary approach to science, assume that exaggeration is our only technique, it is – in fact – merely one of several that we find
serve well in everyday discourse. Exaggeration, like a lever
that can move the earth, hinges on one very important point:
scale.
We, of course, are always meticulous.
One time we came into the lab and there was a spider crawling out the door hauling behind it a silk-lashed package of
documents. We were initially of a mind to crush it, but we realized that interrogation might yield more information about
this rather odd state of affairs. We dropped a spare flask over
the spider and set about fashioning eight tiny foot-cuffs.
Notice how in Figure 1 we have enlarged the left-hand
scale. This is perfectly acceptable behavior, as we are simply
“zooming up” the relevant portion of the chart. (Note: error
bars are not included in this chart because they would exceed
by many times the size of the page. We always recommend
omitting error bars, as they may conflict with your notions of
significance.)
Fifteen minutes later we were getting all up in spidy’s
mandibles demanding answers. We are not frightening to
behold – in fact, many members of the same and opposite
sexes stand rapt as we enter and leave rooms, yearning after
our form – but multiplied into hundreds by faceted eyes and
slobbering something fierce (it being nearly lunchtime), we
must have been quite a sight for the entrapped web-slinger,
for it confessed nearly instantly.
Similarly, many researchers neglect to HFA (home-fieldadvantage) correct their plots. This is inexcusable, as such
a correction is trivial to perform (merely scale your results
by a factor between 0.1x and 10x depending on weather
conditions and “knowing the terrain”), and generally provides a more favorable – i.e. relevant – comparison.Without
HFA scaling, for instance, our box-and-circle plot (Figure 2)
might have been significantly less flattering to our pride.
3.2
We freed a limb so that the arachnid could scrape out its
confession on an aging Ben & Jerry’s receipt with a bit of
graphite we’d located in a dusty corner.
But the spider proved to be more cunning than we had imagined – swiftly imbibing a heretofore concealed cyanide capsule rather than penning the damning testimony we sought.
No matter! For we had recorded the initial exchange and
now had a bargaining token of immense power in our quest
to suppress the nearby lablet that had originated the clandestine agent.
Suppression
Result suppression is a specific form of hyperbolic rendering
that leverages the inadequacy of other researchers to your
advantage. Consider Figure 3 – in this case, we may or may
92
Figure 4: Our system diagram is great, isn’t it? All small
and cute. Probably warm and fuzzy too. Really makes you
want to just take our system home and let it play in a cardboard tube. Or maybe in some wood chips.
Figure 6: If you carefully examine this data without being
distracted you will certainly realize that our method is better
than all others.
wonder if this difference is due to some sort of childhood
trauma involving a continuously collapsing isomorphism and
painful space folds, or an aversion to certain line styles.
However, we leave this last crumb of future work to assure
ourselves another year of funding despite what is about to
transpire in the next section.
Figure 5: Baby, I feel like we got a great connection goin’
on between these two graphs. Like we could get lost in this
moment and it could last a lifetime, with our facets a, b, and
c aligned in delicate union.
4 We Solve Science (a user study)
Indeed, for all that our competitors may have a more theoretically sound algorithm (and we emphasize “may”). They
consort with devil-creatures in order to obtain and research
background material, and for this they should be put to death.
Preferably before any details of their method can be published or otherwise distributed to our peers.
As a final result, we present the results of a user study in
which we solve science entirely. First the raw data:
Solved?
Yes
No
3.3 Distractors
Answers
% Correct
95
Yes
5
No
Notice that while only 95% of respondents agreed that we
had solved science, the remaining 5% were wrong (we are
100% confident of this, of course). Thus, we have solved
science.
Distractors are an emerging phenomena in hyperbolic rendering. They seek to build a virtual bridge between a stimulus (a graph or other figure) and a feeling (warmth, security, curiosity, arousal, etc...). For instance, in Figure 4, we
provide a system diagram that emphasizes how compact and
twitchy and gosh-darn cute our approach is.
Having solved science, this AN602 of a paper has no choice
but to air-burst onto the Novaya Zemlya-like scientific scene,
washing away any Castle Bravo-styled pretenders and remaining an active threat for many years to come.
We caution other researchers that some distractors may have
preference-biased effects, however. For instance, when trying to make a meaningful connection between triangulated
planar graphs (Figure 5), we discovered that reader reaction
varied greatly with background. In this case, we found that
while many subjects (≈ 30%) did report a strong connection between the graphs, others (≈ 20%) simply though it
was going to be more of a one-night isomorphism, and many
readers (≈ 50%) simply weren’t interested.
Acknowledgements
Thanks to flickr users Nesster, vanessa hud, cdrussorusso,
and Idhren for valuable distractions, and to the reviewers for
putting up with this dreck.
Oddly enough, the roughly 50% of our gender-balanced sample that were not interested by the planar graphs were somewhat more titillated by our plotted result data (Figure 6). We
93
ach
sigbovik
Paper 9: Hyperbolic Rendering for Result Enhancment
Reviewer 1
Rating: 2 (accept)
Confidence: 2/4
This paper presents an innovative new strategy for obtaining good results. So many times in my extensive research background I have obtained poor and embarrassing results that one might wonder why
I still have funding. Perhaps it is out of pity. Sometimes I burn my results out of shame, or in a fit of
frustration print them out and tear up the paper to remove them from existence. After having read this
paper I can say with the highest confidence that, with one simple citation back to Mr. N’Ncamc who
pioneered this most excellent tactic, any graphs and tables that may smuggle themselves into my future
work will be forced to do so under the guise of successful data.
There was a picture of a kitten in this paper, as well as a picture of a hamster next to a picture of a
banana. This paper disappointed me somewhat with its lack of development into these two critical matters. I would be interested to see studies in the future as follows: “On the Applicability of using Kittens
as Error-Bars” and “Can the Hamster Get and Eat the Banana? A Cardboard-Tube-Oriented Approach”
so that laypeople may better understand the theory behind this paper.
94
The Box and Circles Plot: a Tool for Research Comparison
John Thomas Longwood∗
Abstract
mute some or all data into spatial dimensions, extracting redundancies directly (as with PCA), with distance-preserving
embeddings [Torgerson 1958], or through iterative methods [Kaski 1997]. Others take a discrete approach, building
space-filling hierarchies [Shneiderman 1991] or unrolling
data along space-filling curves.
Often, when making a survey of work, researchers need to
simultaneously summarize and compare the relative performance, applicability, practicality, novelty, distribution, and
impact of several different pieces of research. We present
new approach – the box-and-circles plot – for the presentation of these high-dimensional data. Our construction, in the
spirit of Chernoff faces, is anatomically motivated. However,
in contrast to previous embeddings, our comparison charts
use a semantic mapping between data and figure, making
it easier for viewers to assign meaning from the differences
they recognize in the final plot.
Other methods, such as our own, seek to maintain all axes in
the data, either by spatially separating them [Inselberg and
Dimsdale 1990], or by assigning them to properties of an
abstract [Pickett and Grinstein 1988] or concrete [Chernoff
1973] figure. Our work fits into the second category, but
is unique in that the assignment is semantically motivated
rather than arbitrary.
CR Categories: I.x.1 [Visualization]: Charts and Graphs—
Multivariate Data
3 The Diagram
Our box-and-circles diagram uses five axes of variation
among three graphical elements. We introduce these elements individually, then show how they can be assembled
into a completed whole. Though one could use these components in isolation, certain subsets may hinder readability.
1 Introduction
We present a new figure, the box-and-circles plot, wellsuited to the summarization and comparison of many of
the relevant variables when presenting a survey of work.
Like Chernoff faces [Chernoff 1973], the legibility of our
proposed diagram relies on anatomical analogy. However,
where Chernoff faces make use of an arbitrary mapping from
dimensions to output perturbations, our box-and-circles plot
relies on a semantic mapping. This provides greater figure
legibility, though does restrict the application domain of our
construction.
3.1
In the remainder of this paper we summarize existing work
on multidimensional visualization; introduce the various
components of a box-and-circles plot and the semantics of
their deformation; provide several examples of completed
plots; and, finally, discuss possible extensions to increase the
usability of our method.
The Box
The primary element of the box-and-circles diagram, as the
name suggests, is a box. It is somewhat like a bar in a bar
chart, but it is significantly more expressive, as we assign it
three degrees of freedom instead of merely one.
2 Background
3.1.1
Box Length: Quality
The effective presentation of high-dimensional data is a
long-standing goal of visualization research. Many general
approaches have been presented, and we briefly mention exemplars from several categories. We refer the astute reader to
examine a recent survey [Keim and Kriege 1996] for further
information.
All high-dimensionality-data visualization systems rely on
some sort of information collapse. Some systems trans∗ e-mail:
The most straightforward variation in our box-and-circles diagram is the bar-chart-like variation of box length. To this
ix@cmu.edu
95
3.2 Circles
axis, we assign the “quality” of the work – that is, the critical variable against which the success of the result under
question is judged. Depending on area, this could be algorithm running time, catalyzed reaction rate, theoretical performance, suppressed angular excursion, increase in agent
tolerance, tensile strength, reduction in infection, manic defiance, butterfly option spread, induced pleasure, tightness of
bound, reflex adjustment, elapsed time, fusion rate, flexible
spend, user efficiency, or something else entirely.
Implementation Note: It is crucial to map the quality variable such that longer boxes indicate better results, as not
doing so would conflict with popular wisdom and possibly
confuse viewers.
3.1.2
The second component of the box-and-circles diagram are
the circles. As they are a less familiar element, we assign
them only two degrees of freedom.
Implementation Note: though – for realism – the right circle
should generally be drawn slightly lower, we recommend –
for readability – drawing the background circle lower.
Box Width: Applicability
3.2.1
Unlike a bar (where width is simply an aesthetic concern),
our box’s width has meaning. When comparing raw performance numbers it is often the case that one should take into
consideration the scope of the approach in question; that is
to say, some approaches perform well but are limited by restrictive assumptions about the problem domain or setting.
As problem scope is often referred to as “breadth” or “girth”,
this seems a natural fit for the box width variation.
3.1.3
Circle Size: Novelty
We vary the radii of the circles in order to denote the audacity of the approach. This is important as it is often useful
to distinguish incremental refinements from totally new approaches, though they may be similar in quality, practicality,
and performance.
Box Bend: Practicality
3.2.2
Though bend is not an attribute commonly associated with
boxes, we find that most viewers are nonetheless able to
distinguish it from both width and length. We use bend to
represent the practicality of a research approach – an upward bend denotes a ready-for-deployment approach while
an downward bend indicates an approach that still requires
further development effort (be it research or engineering) to
be put into practice.
Circle Radial Lines: Distribution
One can add lines to the circles to indicate the amount of
distribution that a concept has achieved. This has both the
visual effect of demonstrating the knowledge radiating outward, and the co-location with the novelty demonstrates the
connection between the two.
Implementation Note: whatever measure is arrived at to
gauge distribution, it is probably reasonable to scale it such
that between zero and twenty lines are drawn. One’s wrist
can become tired otherwise.
Grant-writing Note: we find that bend is particularly useful
in conveying to funding agencies which programs of research
require further stimulation.
96
Figure 1: The impact graph visualizes ideas that stem from a work under consideration. The graph is constructed by (left)
plotting the results of a literature search; (center) creating a rough outline; and (right) removing the extraneous detail. Note
that [Comic 2009] was so inept as to forget to cite related work. Such papers can still be included, generally as disconnected
regions.
3.3
Impact Graph
The final element of our diagram is the impact graph (Figure 1). This is an approximate visualization of the follow-on
work inspired by the idea in question. Its construction begins
with a standard literature search, collecting ideas that seem to
grow from the paper under consideration. These ideas form
a natural graph structure, and so can be laid out as such, with
time making up one axis and the other being used to prevent
overlap. Of course, such a detailed structure is more more
confusing than illuminating, which is why we abstract the
graph by drawing a rough outline. This outline still clearly
shows the progression of an idea through the scientific community.
Figure 2: Integration of Chernoff face into box-and-circles
diagram. The face can be used to convey additional degrees
of freedom, with the down-side that these degrees of freedom
may not have a clear semantic mapping.
Implementation Note: Of course, for recent discoveries, generating such a diagram would be premature; in this case, we
recommend its omission.
3.4
4 Examples
In Figure 3 we provide examples of some of the many variations of the box-and-circles plot in action. Notice how the
plot conveys ideas with a certain immediacy that other methods lack. In user studies many participants found it difficult
to look away from the plots, even though they knew they
should really be concentrating on other features of the text in
question.
Putting It Together
5 Discussion
We have presented an anatomically-motivated figure for the
display of high-dimensional data. In contrast to previous approaches, our figure makes use of semantic mappings between figure and data. While this does limit its applicability,
it also makes the resulting diagram far more readable.
As you can see, the elements of the box and circles plot work
together to form a complete picture of the research in question: (a) box length indicates quality; (b) box width indicates
applicability; (c) box bend indicates practicality; (d) circle
radius indicates novelty; (e) radial lines indicate distribution;
and (f) the impact graph indicates follow-on work.
In the future, we plan to investigate ways to add yet more information to the diagram. One promising avenue for exploration is the integration of Chernoff faces [Chernoff 1973]
without our box-and-circles diagram. This would allow the
addition of extra degrees of freedom – though, unfortunately,
these would not be semantically mapped. We present and
97
Figure 3: An example of three box-and-circles plots. Top, impractical but novel research; not widely distributed, but with a
torrent of follow-on work. Middle, broadly applicable and eminently practical work. More distribution than the first work, but
not as well-cited. Bottom, high-performance solution to a narrow problem subset, using conventional techniques. Distributed
widely, but not much follow-on.
IEEE Visualization.
early result in this direction in Figure 2.
K ASKI , S., 1997. Data exploration using self-organizing
maps.
We are also aware of a certain chromosome-bias in our work.
It may be possible to rectify this through a choice of different
anatomical structures; however, we feel that other options
may lack the significant semantics of characteristics that we
draw upon in our present work.
K EIM , D. A., AND K RIEGE , H.-P. 1996. Visualization
techniques for mining large databases: A comparison.
IEEE Transactions on Knowledge and Data Engineering
8, 923–938.
Finally, we remind researchers that – though our diagrams
may make them feel somewhat inadequate – it’s not the performance, applicability, practicality, novelty, distribution, or
impact of your research that matters; it’s how you use it.
P ICKETT, R. M., AND G RINSTEIN , G. G. 1988. Iconographic displays for visualizing multidimensional data. In
Proceedings of the IEEE Conference on Systems, Man,
and Cybernetics, 164–170.
References
C HERNOFF , H. 1973. The use of faces to represent points
in k-dimensional space graphically. 361–368.
S HNEIDERMAN , B. 1991. Tree visualization with treemaps: A 2-d space-filling approach. ACM Transactions
on Graphics 11, 92–99.
I NSELBERG , A., AND D IMSDALE , B. 1990. Parallel coordinates: A tool for visualizing multidimensional geometry.
T ORGERSON , W. S. 1958. Theory and Methods of Scaling.
Wiley.
98
Department Productivity in Computer Science as a Ratio of
Gender Diversity
Leigh Ann Sudol
Pittsburgh, PA
lsudol@andrew.cmu.edu
ABSTRACT
There has been a large amount of research involving increasing the gender ratio of women in Computer Science
departments. Many people have postulated the effects of
such gender balancing but few papers have been published
with actual resulting statistics of such. In this work we create a model of a computer science graduate department and
through simulation can determine actual impact of varying
gender ratios. First we will demonstrate the givens of such
a simulation system and then move to results of the simulation.
K.3 [Computers & Education]: Computer and Information Sciences Education—Gender Issues
General Terms
Gender, Department Productivity
As all good researchers know, if you can build a model and
assign variables which will then produce graphs, you have
conducted strong research. Additionally this research is easily replicated by any other researcher by creating the same
model, using my stated givens and thereby obtaining the
same results. The model can also be easily generalized by
simply changing the names of the variables to fit any situation you might have.
In the following sections we define the all important system
givens as well as
2.
SYSTEM GIVENS
Assuming that productivity of a department in all of its
intricacies can be quantified, I propose the following metrics
as measures of department productivity.
Keywords
Gender
1.
different things, and different things are good. Although the
exact increase in productivity is most likely variable based
on the characteristics of the individuals, we are claiming it
can be viewed as a constant for the simplicity of our model
and to produce more impactful-looking results.
INTRODUCTION
There has been a very large amount of research and a large
amount of money spent to increase diversity in computer science departments. The National Science Foundation(NSF)
has launched several major initiative costing millions of dollars including the Broadening Participation grant and CCLI
among others. Good tax dollars once again being spent to
support unresearched mandates because people of influence
can yell really loudly.
The work presented in this paper attempts to show an accurate, quantitative analysis of the integration of gender diversity into a prominent computer science department. We
make the assumption that diversity (in the absence of all
other variables) will yield a more productive department;
relying on the little proven fact that different people make
2.1
Paper Output
We all know that the more papers output, the more productive you are. This metric can therefore also scale to a
department level. This is the total count of papers for which
members of the department have published during the time
between black Fridays (our unit of time measurement).
2.2
Quality of Publication
For each publication received by a member of the department in the time unit given, a quality factor will be assigned based upon the number of citations it receives, the
importance of the publication location (ie. papers published
on my fridge are worth less than SIGCHI which are worth
much less than SIGBOVIK), and the number of people who
mention the publication to the author without prompting.
2.3
Number of Talks Given
Assuming most talks do not fit in the Bovik quadrant[1] a
talk should be considered to increase the productivity of the
department.
2.4
Erdos Numbers
Of course the ultimate goal of the department should be to
decrease both the variance and average of the Erdos numbers
99
held by faculty and grad students. Since Erdos numbers
of infinity when combined with countable integers cannot
possibly contribute in a meaningful way to this statistic we
propose reducing infinity to 10000 for the purposes of this
paper.
2.5
Money
Many department chairpersons, unable to secure funding
for an entire department all by themselves, enlist faculty to
do this for them. It is therefore important for department
members to get money, and should be considered part of the
productivity of a department.
2.6
Number of Offices Occupied
This is self evident. I do however divide the offices into those
that hold a female and those that do not for future model
creation.
2.7
Collaboration Quotient
The number of collaborations between members of the department can help impact several of the previously stated
metrics, and so they are given bonus points here.
2.8
Requisite Model Equations
The following equation is put forth as a model for the productivity quotient of the department.
P = Quality ∗ P apers + αT alks + βErdos + γM oney +
δOf f icem + γOf f icew + ζCQ
As shown in [2] the above equation is a viable measure of
department productivity.
2.9
SIMULATION
A simulation was built of the aforementioned variables to
determine that as we increased variables in the equation the
productivity also increased. We also needed to fine tune the
greek constants for the department at CMU. It is recommended that to generalize this to any university department
that those parameters be fine tuned based on the culture of
your department and the particular bias of its members. For
example, if your department contains particular PL professors who feel their research is more important than systems,
then those conferences should be weighted more heavily in
the quality of publication matrix.
3.1
Live Verification
We then ran the simulation on the department by observing
that members of the CS department behaved in predicted
manner. Results of this have been published in psychological
studies such as [3] and [4].
4.
Our simulations on humanity have however indicated anomalies in this model, and the premises of this paper is based
upon a closer examination of those anomalies.
4.1
Social Relationships
There is a particular observed behavior in the members of a
department who make up the majority gender. In computer
science departments that happens to be males (unless it is
a all women university) and so as such we will refer to the
majority gender as male from this point on.
Related Work
None, since no one was willing to make the politically incorrect (but scientifically correct) argument I am about to.
3.
Figure 1: Figure 1:Standard Gender Ratio Curve
IMPORTANT GRAPHS
If we consider only the original equations for productivity
and the aforementioned indicators the graph appears to be
relatively straightforward as shown in Figure 1.
100
It has been observed that for any woman in the computer
science department there is some number of males inhabiting
her doorway or office trying to date her. This number is
dictated by the actual space in the office and can be set to
a constant value of m in the model. Since they are trying
to date her and have therefore stopped doing research there
is an addition variable to be added to the model equation
which represents the decrease in productivity of these men.
Since the graphs are symmetric (as the opposite effect occurs
when women are in the majority) we will look at only the
ratios up to an equal gender division in Figure 2.
Notice in Figure 2 that as the ratio increases to halfway, the
productivity of the department actually decreases since the
increase from diversity finally overtakes the decrease from
either trying to obtain a relationship, or actually trying to
maintain one.
5.
ANALYSIS AND RECOMMENDATIONS
This is ongoing research and as such many more results are
expected to be presented in the actual presentation. The
impact of hallway work ethic for the enticement of future
dating, and occasional ECE student dating (of either gender) as well as the confounds of distance relationships will
be explored as the data emerges.
It is recommended that departments spend careful time fine
Figure 2: Figure 2:Gender Ratio Curve With Desire
to Date
tuning the parameters of their productivity quotient in order to create accurate models of behavior and then use that
model to target a particular population of admitted students to attend the university. Since it is unethical to bias
the admission rates due to gender, it may also be advisable
to create openings in particular departments with an eye towards the historical contribution they make to the gender
ratio.
Overall, smaller offices (for less dating-waiting behavior) or
locks on female office doors may be advisable. Also, when
asked for advice about how to attract a date in advisor meetings, or during social events containing faculty current students should be encouraged to demonstrate their worthiness
as a potential significant other through research and other
methods that would increase the department productivity
quotient.
6.
REFERENCES
1 Vaniea,Kami 2010 ”How to Successfully Prevent the Flow
of Information in Research Presentations”, in Proceedings
SIGBOVIK 2010, Pittsburgh PA
2 Authors, Other 2015 ”A Retrospective Analysis of the
Equation for Department Productivity”, Bovik Press, Pittsburgh PA
3 Bovik, Philomenia, ”Verification of Model Equations on
Biological Entities Existing in a Computing Environment”,
Baker Hall Symposium for the Humanities, 2009
4 Crystal, Billie, ”Understanding Technical Department Productivity in the Wild:How equations can sometimes be scary
but useful”, David Klar Presents SCIENCE! Consortium,
2008
101
102
Track 1
Productivity, We Say Ironically
О
ach
MYTHBUSTERS
Functional Perl: Programming with Recursion Schemes
in Python ........................................................................................105
Robert J. Simmons, Nels E. Beckman,
Dr. Tom Murphy VII, Ph.D.
О
THE BIGGEST LOSER
First Year Student’s Thoughts on Reflections on Bovik’s
Seminal Work: Professional Student’s Strategy for Perpetual
Funding at CMU ............................................................................111
О
IRON CHEF
Feasibility Analysis of Cooking Spaghetti with Meatballs ............113
Ben Blum
О
ROBOT CHICKEN
Cat-a-loging !!11! ..........................................................................115
Leigh Ann Sudol, Ciera Jaspan
О
THE HUMAN TARGET
A novel approach to managing information overload in
syndicated feed readers ..................................................................77
Matt Mets
103
Track 1
Productivity, We Say Ironically
(continued)
О
ach
PIMP MY RIDE (THIS ONE WAS SO OBVIOUS I ACTUALLY FEEL BAD ABOUT IT)
“Pimp My Library” Is An Affront To Pimpers Of The World,
Everywhere ....................................................................................121
Dr. Donna M. Malayeri, PhD
О
FLASH FORWARD
Computation in the Complex Time Plane ......................................123
Matthew Rodriguez
104
Functional Perl: Programming with
Recursion Schemes in Python
Robert J. Simmons
Nels E. Beckman
Dr. Tom Murphy VII, Ph.D.
{rjsimmon,nbeckman}@cs.cmu.edu
The tom7.org foundation
{tom7}@tom7.org
Abstract
tm 1 ⇓ x1
Algorithms that are fundamentally expressions of structural induction over a polynomial data type are famously awkward to
implement in object-oriented programming languages, leading to
three-day-bucket hacks like the “Visitor Pattern.” We show that the
exception-handling mechanism present in most object-oriented languages already suffices.
Conventional wisdom dictates that exceptions are in essence
about the non-local propagation of errors. Upon close scrutiny, we
find the contrary: exceptions are fundamentally about unchecked
depth-1 structural pattern matching.
We give examples of this programming idiom in Python, Perl,
and Java. JavaScript requires a language extension, which we build
and deploy. We then show how two styles by which this style may
be re-integrated into programming languages in the ML family. We
conclude by suggesting design changes that would facilitate the use
of this idiom in other languages.
Techniques]: Standards
int(x) ⇓ x
Using the pattern matching syntax present in the ML family of
languages, the implementation of this big-step evaluation semantics
is similarly straightforward:
fun eval tm =
case tm of
Int x => x
| Plus (tm1, tm2) => eval tm1 + eval tm2
We can also similarly implement a small step semantics for this
language as defined by the following rules:
int(x) value
tm 1 value tm 2 → tm 2
+ tm 2
tm 1 + tm 2 → tm 1 + tm 2
x 1 + x2 = x3
int(x1 ) + int(x2 ) → int(x3 )
The ML implementation is a straightforward implementation
of these inference rules, with the exception of the need to define
an additional cast function that we call to get the value of the
enclosed integers once we know the terms to be a values.1
tm 1 →
tm 1 + tm 2 →
D.2.3 [Coding Tools and
General Terms Algorithms, Design, Languages, Standardization
Keywords pattern matching, recursion schemes, structural induction, python, perl, java, javascript, types
1.
tm 2 ⇓ x2 x1 + x2 = x3
tm 1 + tm 2 ⇓ x3
Introduction
Pattern matching is a convenient way of representing polynomial
datatypes, which can be informally thought of as a generalization
of tree-like data. An extremely simple example is an abstract syntax
for a language with integers and addition:
tm 1
tm 1
exception Stuck
fun value tm =
case tm of
Int x => true
| _ => false
datatype tm =
Int of int
| Plus of tm * tm
fun cast tm =
case tm of
Int x => x
| _ => raise Stuck
The big-step evaluation semantics for this language is quite straightforward:
c 2010 by Robert J. Simmons, Nels E. Beckman, and Tom Murphy VII.
Copyright Permission is hereby granted, free of charge, to any person obtaining a copy of this
software and associated documentation files (the ”Software”), to deal in the Software
without restriction, including without limitation the rights to use, copy, modify, merge,
publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons
to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies
or substantial portions of the Software.
THE SOFTWARE IS PROVIDED ”AS IS”, WITHOUT WARRANTY OF ANY
KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE
AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR
THE USE OR OTHER DEALINGS IN THE SOFTWARE.
fun step tm =
case tm of
Int x => raise Stuck
| Plus(tm1, tm2) =>
if value tm1
then if value tm2
then Int(cast tm1 + cast tm2)
else Plus(tm, step tm2)
else Plus(step tm1, tm2)
1 In
a language without static typing and with distinct varieties of values,
these casts can, of course, fail if we get the wrong variety of value.
105
class tm:
pass
class Int(tm):
def __init__(self, x):
self.x = x
class Plus(tm):
def __init__(self, e1, e2):
self.e1 = e1
self.e2 = e2
interface tm
{ public <T> T accept(Visitor<T> visitor); }
class Int implements tm {
int x;
public Int(int x) { this.x = x; }
public <T> T accept(Visitor<T> v)
{ return v.visit(this); }
}
class Plus implements tm {
tm e1;
tm e2;
public Plus(tm e1, tm e2) {
this.e1 = e1;
this.e2 = e2;
}
public <T> T accept(Visitor<T> v)
{ return v.visit(this); }
}
def eval(tm):
try: raise tm
except Int: return tm.x
except Plus: return eval(tm.e1) + eval(tm.e2)
def value(tm):
try: raise tm
except Int: return True
except: return False
interface Visitor<T> {
public T visit(Plus p);
public T visit(Int i);
}
def step(tm):
try: raise tm
except Plus:
if(not value(tm.e1)):
return Plus(step(tm.e1), tm.e2)
elif(not value(tm.e2)):
return Plus(tm.e1, step(tm.e2))
else: return Int(tm.e1.x + tm.e2.x)
public class VisitorExample {
static int eval(tm tm) {
return tm.accept(new Visitor<Integer>(){
public Integer visit(Int i) { return i.x; }
public Integer visit(Plus p) {
return p.e1.accept(this)
+ p.e2.accept(this);
}
});
}
ex_1 = Plus(Plus(Plus(Int(1),Int(2)),
Int(3)),Int(4))
print ("Evaluating: Plus(Plus(Plus(1,2),3),4) - "
+ str(eval(ex_1)))
print ("Stepping x3: Plus(Plus(Plus(1,2),3),4) - "
+ str(step(step(step(ex_1))).x))
public static void main(String[] args) {
tm ex_1 =
new Plus(new Plus(new Plus(new Int(1),
new Int(2)),
new Int(3)), new Int(4));
System.out.println
("Evaluating: Plus(Plus(Plus(1,2),3),4) - "
+ eval(ex_1));
}
Figure 2. Python pattern matching case study.
mechanism, which Java also incorporates, is fundamentally about
unchecked depth-1 structural pattern matching, we can seek to use
exception handling to provide a similar sort of depth-1 structural
pattern matching to the Java language and other object-oriented
programming languages. The rest of the paper explores this possibility.
}
Figure 1. Case study implemented with the Visitor Pattern in Java.
2.
While there is a somewhat more pleasant way to implement the
Step function using nested pattern matching, we will only consider
the “shallow” pattern matching we see here in this paper.
1.1
Discovery and implementation in Python
Imagine we have the abstract syntax of our example encoded in
Python in the standard way, as a class hierarchy where subclasses
Int and Plus extend the basic abstract syntax class tm.
class tm:
pass
class Int(tm):
def __init__(self, x)
self.x = x
class Plus(tm):
def __init__(self, e1, e2)
self.e1 = e1
self.e2 = e2
Object-Oriented languages and the Visitor Pattern
While ML-family languages suggest describing different types of
syntax as different branches of a subtype, most object oriented
languages suggest describing different types of syntax as different
objects that either implement the same interface or extend the same
subclass. The traditionally understood way of performing elegant
recursion over these structures is known as the Visitor Pattern; the
implementation of big-step evaluation using the Visitor Pattern can
be seen in Figure 1.
The Visitor Pattern, whatever its merits may be, is certainly
quite different in character from the depth-1 pattern matching used
in our ML example. However, because the exception handling
The big-step evaluation semantics of this language could be
written by explicitly using isinstance tests, but the repeated use
of elif isinstance(tm, ...) is bulky and unsatisfying, which
only becomes more true when the code has many, many branches:
106
class tm extends RuntimeException {}
final class Int extends tm {
final Integer x;
Int(Integer x) { this.x = x; }
}
final class Plus extends tm {
final tm e1;
final tm e2;
Plus(tm e1, tm e2) {
this.e1 = e1;
this.e2 = e2;
}
}
def eval(tm):
if isinstance(tm, Int):
return tm.x
elif isinstance(tm, Plus):
return eval(tm.e1) + eval(tm.e2)
else: raise RuntimeError
Note the final line of the above code, which raises a runtime
error if the evaluated term is neither an Int nor a Plus. This is
important in case we extend the abstract syntax tree with more
branches; we want something equivalent to the “nonexhaustive
match exception” in an ML-family language.
2.1
Utilizing the exception handling mechanism
public class PatternMatch {
The first key observation for introducing our idiom is the fact
that every single object in Python can be treated as an exception
and thrown; the language designers suggest that exceptions derive
from the Exception class, but there is no enforcement of this
mechanism. The second key observation is that the branches of
an exception handling call can check the object’s class tag when
deciding which branch of the exception handling to consider. The
motivation is to allow, say, a divide-by-zero error or a user-input
error to be both caught at the same place but handled in different
ways; we can use it, however, to do exactly the isinstance test
from before in a more concise, pleasing notation:
static Integer eval(tm tm) {
try{throw tm;}
catch(Int i) { return i.x; }
catch(Plus p)
{ return eval(p.e1) + eval(p.e2); }
}
static Boolean value(tm tm) {
try{throw tm;}
catch(Int i) { return true; }
catch(Plus p) { return false; }
}
def eval(tm):
try: raise tm
except Int: return tm.x
except Plus: return eval(tm.e1) + eval(tm.e2)
static tm step(tm tm) {
try{throw tm;}
catch(Plus p) {
if( !value(p.e1) )
return new Plus(step(p.e1), p.e2);
else if( !value(p.e2) )
return new Plus(p.e1, step(p.e2));
else
return new Int
( ((Int)p.e1).x + ((Int)p.e2).x );
}
}
Furthermore, this form of evaluation introduces a natural notion
of a nonexhaustive match exception: if we add new kinds of terms
(say, a multiplication) and don’t handle them in the eval function,
the very abstract syntax tree node that we did not handle will be
thrown as an exception: an error that potentially contains much
more helpful information than an ML Match exception!
2.2
Adaptation in Java
The relatively-more-strongly-typed language Java also allows for
the expression of our idiom by simply allowing every term to extend the class RuntimeException. Technically, all that is necessary for the use of our idiom is Exception, but by the use
of RuntimeException we avoid dealing with compiler warnings
caused by lack of exhaustiveness checking (similar to the use of the
MLton SML compiler’s -disable-ann nonexhaustiveMatch
option).
Our case study can be seen implemented in Java in Figure 3. It is
particularly critical that, in Java, the exception handling mechanism
handles appropriately casting the term tm to an integer Int i or
an addition Plus p. The typed nature of Java makes the repeated
instanceof tests much more painful than they were in Python,
because we must not only do repeated if-then-else statements but
also downcast terms in branches where we statically know that
the cast is safe already as a result of the instanceof test. The
exception-handling idiom is free of this particular annoyance.
3.
public static void main(String[] args) {
tm ex_1 =
new Plus(new Plus(new Plus(new Int(1),
new Int(2)),
new Int(3)), new Int(4));
System.out.println
+ eval(ex_1));
}
}
Figure 3. Java pattern matching example.
3.1
JavaScript
JavaScript is a curly-like language oriented on the concept of
JavaScript objects, which are functions [1]. The typical way to implement pattern matching would be to use a series of instanceof
tests, like in Java or Python. As before, we’d like to improve this
code to something like what appears in Figure 4.
Unfortunately this does not work because it is not legal JavaScript
code. It doesn’t parse or run or anything. A typical blunder:
JavaScript denies the nature of exceptions by only allowing them to
Implementation via language extension
Some languages do not have the essential aspect required for case
analysis to be implemented in the exception-handling mechanism.
We describe an extensive effort to implement this functionality
in JavaScript, as well as commenting on a language where that
extensive effort had already been done: Perl.
107
var Int = function(i) { this.i = i; };
var Plus = function(tm1, tm2) {
this.tm1 = tm1; this.tm2 = tm2;
};
Function.prototype.enhance = function() {
var s = ’’ + this;
var name;
if (s.indexOf(’function (’) == 0) {
s = ’function $(’ + s.substr(’function (’.length);
name = ’$’;
} else {
name = /function *([a-zA-Z0-9_$]+) *\(.*/m.
exec(s)[1];
}
var s = ’(function(){’ + rewrite(s) +
’\nreturn ’ + name + ’})()’;
return eval(s);
};
function big_eval(tm) {
try {
throw tm;
} catch (Int(e)) {
return e;
} catch (Plus(e)) {
return new Int(big_eval(e.tm1).i +
big_eval(e.tm2).i);
} catch (e) {
throw ’stuck’;
}
}
A function value’s source code might be a function declaration
or a function expression.2 We put it in a normalized declaration
form, giving the function the name $ if it doesn’t have one. We
wrap the function declaration in another function and call it immediately, because functions are the only way3 to introduce new
scopes in JavaScript (here thankfully the designers understood that
the essence of functions is scope delineation).
The bulk of the language extension is implemented in the recursive rewrite function. The function rewrites a source code string;
the source code can represent a declaration or an expression or anything. Using strings to represent data helps us avoid tricky type errors that can arise when there’s a difference between expressions
and declarations. The implementation of rewrite appears in Figure 7.
The rewrite function parses the source code to find appearances
of the catch keyword, then takes any with blocks that appear
after them, and rearranges the code just so. It correctly handles the
case that uses of the extension are nested, as well as recursivelyenhanced functions. The details of the implementation are subtle,
mainly having to do with fiddly bits. An example may help the
reader; the big-step evaluation from Figure 5 is translated to the
code in Figure 7.
Figure 4. Desired JavaScript code.
function big_eval(tm) {
try {
throw tm;
} catch (e) {
throw ’stuck’;
} with (Int(e)) {
return e;
} with (Plus(e)) {
big_eval(e.tm2).i);
}
}
Figure 5. Big-step evaluator, using the language extension.
be used for the non-local propagation of errors, requiring a different
language feature to be used for pattern matching (instanceof). In
many languages we would be at an impasse. Luckily, the authors
of JavaScript had the foresight to foresee the need to improve the
language; it is possible for us to build language extensions within
our programs. We can create a full-featured exception mechanism
that we can then use to encode our programming technique.
The extension works by co-opting the rarely-used with keyword to create a new iterated, pattern-matching try . . . catch construct. The above big-step evaluation code can then be written as in
Figure 5. This is basically the same as in the previous languages,
except as a stylistic choice the required default case is written first
(it is legal for it to be blank or to re-raise the same exception).
You can’t run this code because it is not legal JavaScript. If you
try to run it JavaScript will be like, huh? because this is not what
the with keyword does. To run the code you must first enhance the
big_eval function using our language extension:
It is worth noting in a smaller font that the language extension does not handle the
case that source code appears inside string or regular-expression literals. Another way
to put this is that the language extension successfully ferrets out not just uses of the
new feature but mentions of it as well.
3.2
Perl
Because of the pun on “Functional Pearl” we really wanted to
implement this in Perl, but none of the co-authors really wanted
to learn Perl and we couldn’t find Jason [4].
However, while Perl lacks the instanceof-checking exception
capability that is critical to our purposes, there is a Perl module,
Error.pm, that appears to enrich Perl with the necessary features
by a strategy similar to the one we investigated for JavaScript.4 Unfortunately the name of this module also describes what happens
when the authors tried to use it.
4.
var tm = new Plus(new Int(5), new Int(2));
var f = big_eval.enhance();
alert(f(tm).i);
Re-incorporation in ML
Of course, the beautiful synthesis of exception handling and pattern matching can be ported back into Standard ML. Standard ML
2 It
The enhance property is added to every function by our language extension. It enhances the semantics of the language to add
the iterated try. . . catch. . . with construct. Running enhance returns a new function that works right.
The extension works by converting the function value to source
code, modifying the source code, and then evaluating the source
code to get a new function value.
might also be a native function, in which case we cannot enhance it.
This is okay, because native functions probably did not use our language
extension.
3 One can also use the with keyword but that might mess up the language
extension, like if someone tries to enhance an already-enhanced function,
or the enhancer function itself.
4 http://search.cpan.org/ shlomif/Error-0.17016/lib/Error.pm
108
function next(s) {
var depth = 0;
for (var i = 0; i < s.length; i++) {
if (s[i] == ’{’ || s[i] == ’(’) {
depth++;
} else if (s[i] == ’}’ || s[i] == ’)’) {
depth--;
if (depth == 0) return { p: s.substr(0, i + 1),
s: s.substr(i + 1) };
if (depth < 0) throw ’parse error’;
}
}
throw ’parse error 2’;
}
function rewrite(s) {
var out = ’’;
for(;;) {
var pos = s.indexOf(’catch’);
if (pos >= 0) {
pos += ’catch’.length;
out += s.substr(0, pos);
s = s.substr(pos);
var parens = next(s);
var body = next(parens.s);
out += parens.p;
var def = body.p;
out += ’{\nif(0);’;
for (;;) {
s = /[ \n\t]*([\s\S]*)/m.exec(body.s)[1];
if (s.indexOf(’with’) == 0) {
var s = s.substr(’width’.length);
var pp = next(s);
var body = next(pp.s);
var ce = pp.p.substr(1, pp.p.length - 2);
var ctor = ce.substr(0, ce.indexOf(’(’));
var e =
next(ce.substr(ce.indexOf(’(’))).p;
out += ’ else if (’ + e + ’ instanceof ’ +
ctor + ’) ’ + rewrite(body.p);
s = body.s;
} else {
break;
}
}
out += ’ else ’ + rewrite(def) + ’}’;
} else {
out += s;
break;
}
}
return out;
}
Figure 6. The rewrite function, which implements the bulk of
the language extension.
includes a single extensible datatype, named exn, which also happens to be the type of ML exceptions. A typical elegance: ML emphasizes the dual purpose of exceptions as both a mechanism for
non-local error propagation and for pattern matching by using the
exception keyword to extend the exn datatype:
(function(){function big_eval(tm) {
try {
throw tm;
} catch (e){
if(0); else if ((e) instanceof Int) {
return e;
} else if ((e) instanceof Plus) {
big_eval(e.tm2).i);
} else {
throw "stuck";
}}}
return big_eval})()
Figure 7. The automatically-enhanced version of Figure 5.
structure PatternMatch = struct
exception Int of int
exception Plus of exn * exn
fun eval tm =
(raise tm) handle
Int x => x
fun value tm =
(raise tm) handle
Int x => true
| _ => false
fun cast tm =
(raise tm) handle
Int x => x
fun step tm =
(raise tm) handle
Plus(tm1, tm2) =>
if value tm1
then if value tm2
then Int(cast tm1 + cast tm2)
else Plus(tm, step tm2)
else Plus(step tm1, tm2)
val main =
let val ex_1 =
Plus(Plus(Plus(Int 1,Int 2),Int 3),Int 4)
in
print
^ Int.toString (eval ex_1) ^ "\n\n");
print
("Stepping x3: Plus(Plus(Plus(1,2),3),4) - "
^ Int.toString
(cast(step(step(step(ex_1)))))
^ "\n\n")
end
end
Figure 8. Re-implementation into SML.
The SML parser requires that (raise tm) be surrounded by
parenthesis, but otherwise the code retains its essential character,
exception Int of int
exception Plus of exn * exn
109
Acknowledgments
as shown in Figure 8. Of course, because the extensible datatype
exn is itself amenable to standard case-analysis, we could instead
define our syntax as branches of the exn as we do in Figure 8 and
still do a regular pattern-match on it:
Milo Polte was instrumental in the discovery of this idiom by
encouraging the first author to learn Python. Neel Krishnaswami
was the first person who observed that this discovery was probably
ridiculous, and Noam Zeilberger reminded the first author what orpatterns are and how they work.
fun eval tm =
case tm of
Int x => x
References
[1] ECMA. ECMAScript language specification. Technical Report ECMA262, ECMA, December 1999.
[2] N. R. Krishnaswami. Focusing on pattern matching. In POPL
’09: Proceedings of the 36th annual ACM SIGPLAN-SIGACT symposium on Principles of programming languages, pages 366–378,
New York, NY, USA, 2009. ACM. ISBN 978-1-60558-379-2. doi:
http://doi.acm.org/10.1145/1480881.1480927.
[3] F. Pfenning and C. Schürmann. System description: Twelf - a metalogical framework for deductive systems. In H. Ganzinger, editor,
CADE, volume 1632 of Lecture Notes in Computer Science, pages 202–
206. Springer, 1999.
[4] J. Reed. Gd and l - systems. The Perl Journal, 3(1), 1998.
[5] R. J. Simmons. Functional Perl: Recursion schemes in Python (extended case study). Documented code example(s). ACH-BOVIK-2010003, Association for Computational Heresy, Pittsburgh, Pennsylvania,
2010.
There are two problems with this approach, however. First, in
the case of an unhandled piece of syntax we get an uninformative
Match exception instead of the offending piece of syntax itself.
Second, we because SML attempts to do exhaustiveness checking
on case analysis, every case analysis on an exn that does not have
a catch-all case runs afoul of the static exhaustiveness checking.
Getting a lot of non-exhaustive match warnings when compiling
doesn’t bother some people, but it can be annoying [3].
5.
Conclusion and future work
There is much work to be done in exploring the beautiful synthesis
of exception-handling and pattern matching. For instance, a serious shortcoming of our exception-handling case analysis is that it
does not allow nested case analysis, which suggests a new language
feature – nested exception handling – for Java/Python family languages.
Similarly, the exception-handling idiom allows Java/Python
family languages to do a particularly crazy duck-typing-ish thing.
Say I want to just enumerate all the integer subterms of my expression, which is extended from the example in the paper to include
Plus, Minus, Times, and Div. As long as I called the left hand
subterm e1 and the right-hand subterm e2 across these different
terms, I can write this code very concisely in Python:
def subterms(tm):
try: raise tm
except Int: print "Subterm: " + str(tm.x)
except (Plus, Minus, Times, Div):
subterms(tm.e1)
subterms(tm.e2)
ex_1 = Plus(Times(Div(Int(1),Int(2)),
Int(3)),Minus(Int(4),Int(9)))
subterms(ex_1)
Similarly, in Java this could be done by making all the binary
operations extend a tmBinary class that, in turn, extends the tm
class. In JavaScript, we can continue our approach of implementing
Java- and Python-like extensions via syntactic rewrites and then use
them in an idiomatic way to simulate a native pattern matching feature. We might think to ask that such functionality be incorporated
into ML family languages, were it not for the fact that it’s already
there by way of one of the wackier SML-NJ-specific extensions
of Standard ML, “or-patterns,” which apparently aren’t any more
insane than focusing was already [2].
fun subterms tm =
case tm of
Int x =>
print("Subterm: " ^ Int.toString x ^ "\n")
| (Plus(t1,t2) | Times(t1,t2) | Mult(t1,t2)
| Div(t1,t2)) => (subterms t1; subterms t2)
For a larger code example (the implementation of the static and
dynamic semantics of Plotkin’s PCF), see our extended technical
report [5].
110
First Year Student’s Thoughts
on Reflections on Bovik’s Seminal Work:
Professional Student’s Strategy for Perpetual
Funding at CMU
School of Computer Science, Carnegie Mellon University, USA
abalacha@cs.cmu.edu, yangp@cs.cmu.edu, wolf@cs.cmu.edu
It’s going to work [1] [2]. 1
References
1. Harry Q. Bovik. Professional student’s strategy for perpetual funding at cmu. Technical
Report CMU-CS-93-000, Carnegie Mellon University, May 1993.
2. Dean Sutherland and Maverick Woo. Reflections on bovik’s seminal work: Professional student’s strategy for perpetual funding at cmu. In A Record of The Proceedings of SIGBOVIK
2009, page 3, 2009.
1
hopefully
111
112
Feasibility Analysis of Cooking Spaghetti with
Meatballs
2010.04.01
1
Abstract
In this paper we study the major contributing factors to the possibility of making spaghetti with meatballs
for dinner tonight. Several different environmental factors are considered, and overall benefit weights are
assigned to each, so that results may be comprehensively analyzed for a final result. We studied the available
ingredients in the kitchen, and for those required ones that happen to be missing, further analysis is performed
to determine whether or not a substitute can be provided from those available or if a trip to the grocery
store must be taken. Furthermore, we measure the amount of available time this afternoon and evening,
and determine the amount of which we would need to set aside to perform cooking tasks, assuming optimal
parallelization. Halfway through the investigation, however, it was determined that we actually wanted
chinese takeout, and so efforts were abandoned.
113
114

!
"
#
!
#
$%
&
'
!
(
))
!
*
+%,#+-..)
"
#$
%$&
'(
115
"
!$
",
'$
(
"-
&
"
$
'$
(
'*
"/
&
$
0
$
(
$&
"+
'$
(
116
"6
'&&*
7*($0
+
'&&*
/!/
,
011
1&2
!
'
011
1&#342
'
$011
/
1
117
118
A novel approach to managing information
overload in syndicated feed readers
Matt Mets
March 21, 2010
Abstract
With the rapid expansion of digital news services, and the trend for
users to select content personally, sorting through the available content
quickly becomes a major hangup for consumption of content. A user can
quickly find themselves overwhelmed, and unable to keep up with the
ever-increasing number of news items available.
People suffer from information overload when dealing with their online
content, however they are perfectly capable of handling high-information
video games.
In this paper, I present a novel method for dealing with this situation.
Reaching to the entertainment industry for inspiration, this new method
presents news items as entities in a virtual space, and allows the user to
quickly deal with them using a simple set of familiar tools, including a
laser gun, and star bomb.
119
120
i1JNQ .Z -JCSBSZu *T "O "ŀSPOU 5P
1JNQFST 0G ǔF 8PSME &WFSZXIFSF
%S %POOB . .BMBZFSJ 1I%
"CTUSBDU
*U JT UP UIJT MBTU VTBHF PG 8QUX UIBU XF PCKFDU "
QIZTJDBM -JCSBSZ NBZ CF QJNQFE CZ NFBOT PG DPM
PSGVM MPHPT -$% TDSFFOT -&% IFBE BOE UBJM MJHIUT
SBDJOH DBS TFBUT DISPNF SJNT B TQJOOJOH CPXM
JOH CBMM B QPQDPSO NBDIJOF B DMPUIFT ESZFS JO UIF
CBDLTFBU BOE TP PO #VU IPX DBO TVDI UIJOHT CF
BEEFE UP B EJHJUBM DPNQVUFS TPGUXBSF -JCSBSZ
*OTQJSFE CZ UIF BNB[JOH TVDDFTT PG .57T i8QUX
.Z 3JEFu <> UIF USBOTJUJWF WFSC GPSN PG iQJNQu
IBT TQSFBE BDSPTT UIF *OUFSOFU 4BEMZ UIF PSJHJOBM
NFBOJOH PG UIF QISBTF IBT OPX CFDPNF EJTUPSUFE
UP UIF QPJOU UIBU JU JT B TFSJPVT BŀSPOU UP UIF "NFSJ
DBOBEJBO MBOHVBHF
4XJTT "SNZ ,OJGF -BOHVBHF 6TFT
1SFMJNJOBSJFT
0O 0DUPCFS 1SPGFTTPS .BSUJO 0EFSTLZ
IFSFBGUFS SFGFSSFE UP BT 4DBMJOWFOUPS
JO
GBNPVTMZ
TVHHFTUFE <> UIBU +BWB™ BOE 4DBMB&1'- MJCSBSJFT CF
iQJNQFEu UISPVHI B NFDIBOJTN UIBU BMMPXT BEEJOH
OFX NFUIPET UP FYJTUJOH DMBTTFT 4BEMZ UIJT JT BO
IVHFBTT NJTVTF PG 8QUX UIBU IBT DBVTFE DPOTJEFS
BCMF DPOGVTJPO JO UIF "NFSJDBOBEJBO MBOHVBHF <>
/PX 4DBMJOWFOUPS CFJOH B (FSNBO 8IJUF %VEF
DBOOPU CF GBVMUFE GPS IJT MBDL PG LOPXMFEHF BCPVU
3BQ $VMUVSF BO "NFSJDBO *OWFOUJPO PS %JTDPWFSZ
EFQFOEJOH PO POFT QPJOU PG WJFX
#VU ZFU XF DBO
UPUBMMZ ZFMM BU UIF CMPH QPTU GPS CFJOH JHOPSBOU PG UIF
USVF NFBOJOH PG 8QUX
"DDPSEJOH UP NZ 04 9 4OPX -FPQBSE <> EJDUJPOBSZ
iQJNQu IBT UIF GPMMPXJOH NFBOJOH
WFSC < JOUSBOT > <PGUFO BT O > QJNQJOH
BDU BT B QJNQ
6TFT
• 1JNQ .Z 4FBSDI <>
• 1JNQ .Z4QBDF <>
• 1JNQ .Z 4OPXNBO <>
• 1JNQ .Z #SJEF <>
4QFDJêDT
• 1JNQ .Z -JCSBSZ < >
ǔF 0SJHJOBM 1FDVMJBS 1PTU <> IFSFBGUFS SFGFSSFE UP
BT 011 JOUSPEVDFT 4DBMB QSPHSBNNFST UP B QSP
HSBNNJOH QBUUFSO XIFSFCZ OFX NFUIPET DBO CF
BEEFE UP FYJTUJOH DMBTTFT /PX UIFSF BSF B OVN
CFS PG %PDUPST PG 1IJMPTPQIZ UIF BVUIPST JODMVEFE
UIBU êOE UIJT UP CF B VTFGVM BOE XBZDPPM QSPHSBN
NJOH MBOHVBHF GFBUVSF <> #VU EPFT JU SJTF UP UIF
MFWFM PG 8QUX 8F UIJOL OPU
*O UIF 011 XF TFF UIF GPMMPXJOH VTF PG OPUSFBMMZ
1JNQJOH
0CKFDUJPOBCMF 6TFT
4BEMZ QPQVMBS VTFT PG UIF UFSN GBS TIPSU PG UIF PSJH
JOBM NFBOJOH PG UIF UFSN 8F CFMJFWF UIF NPTU FHSF
HJPVT FYBNQMF JT i1JNQ .Z -JCSBSZu
ǔFSF BSF UXP TFOTFT JO XIJDI B -JCSBSZ NBZ CF
8QUXML JU NBZ CF TPME JOUP QSPTUJUVUJPO PS JU NBZ
CF QJNQFE PVU ǔFSF BSF BMTP BU MFBTU UXP TFOTFT
PG -JCSBSZ IFSF JU NBZ SFGFS UP B QIZTJDBM TUSVDUVSF
FH <>
PS UP EJHJUBM DPNQVUFS TPGUXBSF
/P KPLFMPPL JU <> VQ ZP
ZFBI ZPV LOPX NF
121
cL
DLD
y
Rdy
xy D
NC
vyD
'JHVSF 5SVF -JCSBSZ 8QUXQVO UISPVHI WBMJE 4DBMB DPEF BOE JOWBMJE GPOUT
ʰ ſɨř ɩř ɪƀ
ʰ ſɫř ɬř ɭƀ
ʰ
$PODMVTJPOT
4P OPX UIF NFBTMZ ŜŜ TVQQPSUT BO
NFUIPE "SHVBCMZ ŜŜ JT TP
DPWBSJBOU BOE MBNFBTT UIBU BOZ BEEJUJPO UP JU NJHIU
XFMM CF DBMMFE 8QUXQVO JU PVU #VU UIF BVUIPST BT
XFMM BT BMM NFNCFST PG UIF *OUFSOBUJPOBM $PBMJUJPO
GPS UIF 1SPUFDUJPO PG 1JNQT BOE 1JNQJOH *$111
CFMJFWF UIBU 9[JCJU 2 "MFY *TI .BE .JLF #JH
%BOF 4IBF FU BM XPVME CF BQQBMMFE BU UIF BNPVOU
PG 8QUXQVO SFNBJOJOH UP CF EPOF
3FBM EJHJUBM DPNQVUFS TPGUXBSF -JCSBSZ 8QUXQVO
NVTU JODMVEF BU UIF WFSZ NJOJNVN UIF VTF PG BU
MFBTU POF EJOHCBU "EEJUJPOBMMZ TBJE EJOHCBUT NVTU
CF CBEPOLBEPOLJDBM
3FGFSFODFT
<> ŜŜŵ
<> 04 9 4OPX -FPQBSE WFSTJPO <> ŜŜŵ
-FU UIF 3FBM 8QUXQVO #FHJO
<> ŜŜŵ
<> ŜŜŵ
<> ŜŜŵţʰɏɪɨɰɫ
ǔJT QBQFST OPWFM EJTDPWFSZ JT UIBU UIF DPNCJOBUJPO
PG 6OJDPEF 5PNCBUT <> BOE 4DBMB <> BMMPX VT UP
USVMZ 8QUX PVS -JCSBSJFT
'PS JOTUBODF UP 8QUX ŜŜ XF
NBZ XSJUF UIF DPEF JO 'JH GPOU FOMBSHFE GPS UIPTF
IBSE PG SFBEJOH
"GUFS UIPTF EFêOJUJPOT XF NBZ XSJUF UIF GPMMPX
JOH QSPHSBN
<> ŞŞŜŜŵ
<> śŵŵŜŜŵŵŜţ
ʰɨɮɰɮɭɭ
<> "CPVU SFTVMUT TBZT (PPHMF
<> ŜɮŜŵɰɯŜ
<> ŜŞŜ
<> .BMBZFSJ %POOB $PEJOH 8JUIPVU :PVS $SZTUBM #BMM 6OBO
UJDJQBUFE 0CKFDU0SJFOUFE 3FVTF 1I% UIFTJT $BSOFHJF .FM
MPO 6OJWFSTJUZ 1JUUTCVSHI 1" 64" NxvR
"Pimp Me Out"
ǔJT DPEF BDUVBMMZ XPSLT OP KPLF
0L TP ZPVE IBWF UP QVU JO UIF DIBSBDUFST JOUP TPNF VOVTFE
6OJDPEF DPEF QPJOUT BOE MPBE VQ UIF GPOU CVU UIBUT KVTU JNQMF
NFOUBUJPO EFUBJMT BOE GVUVSF XPSL "MUFSOBUJWFMZ ZPV DPVME KVTU
DIBOHF UIF GPOU PG TFMFDU DIBSBDUFST XIJDI JT XIBU * EJE
3VOOJOH UIJT QSPHSBN JO UIF 4DBMB JOUFSQSFUFS ZJFMET
#JUDIFT
122
Computation in the Complex Time Plane
Matthew Rodriguez
School of Computer Science
Pittsburgh, PA, 15213
March 20, 2010
Abstract
For a long time, computer scientists have been stuck on certain problems that seemed to be impossible to solve efficiently, or even at all. More
frustratingly, some of them were inversifications of trivially easy problems,
such as factoring to multiplication. In an attempt to solve these problems,
uncomputation was developed. It happened on a fortuitous day when a
clever programmer (that is to say, I) realized that by simply reversalating
a program, it could be run backwards, and its inputs could be deduced
from its outputs in the same time as it would take to run the program
forward.
Up until this paper, I had kept uncomputation a secret, but now that
I am finished my research into the area, I am ready to unleash it upon
the world. It is with great pleasure that I present the exciting new field
of uncomputation. As this paper will demonstrate, uncomputation is a
revolutionary new field of computation that will enable us to do amazing
feats that were previously unfeasible, and with incredible speed.
1
The Basics of Uncomputation
As you know, there exists a trivial algorithm to multiply two n-bit numbers in
O(1) time. Just use the * operator and let the deep ECE magic inside your
computer do the rest. By simply taking the reverse of this algorithm, we can
now unmultiply (or factor if you insist), two numbers in O(−1) time. The
minus, of course, comes from the fact that the algorithm is running backwards
and concludes before it begins. A simple implementation in C is given below.
}
;b * a nruter
{ (b tni ,a tni) ylpitlum tni
Those who are not used to writing unprograms may not be able to see it,
so I would like to point out that there is a tab at the end of the second line for
123
unreadibility.
I have found many other uses for uncomputation. One such use is undelete.
If you think this simply lets you recover files you have deleted, you are sorely
underestimating it. You can undelete anything. Some examples are below.
undelete
undelete
undelete
undelete
meaning_of_life.txt
upcoming_winning_lottery_numbers.txt
grand_unified_theory.tex
lost_works_of_shakespeare.zip
Not all of uncomputation is useful. For example, there’s never any real desire
to unsort a list. Uncomputation is not meant to completely replace computation,
only to act as a supplement. Still, undelete alone obsolesces so much of existing
computer science and other forms of uncomputation that I will conclude here
and proceed to the next topic.
2
Imaginary Computation
Let there be a problem Π with two implementations,
√ Q and Q̂, such that Q
has running time O(n) and Q̂ has running time O( n). Q̂ is an optimization
of Q. If we simply negate
√ these algorithms, we get −Q and −Q̂, with running
times O(−n) and O(− n) respectively. Nothing surprising so far. However, if
ˆ with
we first negate Q√and then optimize
it, we get a different algorithm, −Q,
√
running time O( −n), or O(i n). This algorithm is equivalent to iQ̂. As the
running time indicates, it runs in imaginary time, and we must now generalize
our concept of time to include the imaginary time axis.
How does this work in the real world where time√only flows from past to
future? It’s simple, really. The real component of i n is 0, so it concludes
instantly from your point of view. What’s really going on is that the computation is started by your imaginary past self and runs for a certain amount of
imaginary time before concluding with you, located at (now, imaginary now) in
the complex time plane. You, in turn, begin the computation and it proceeds
toward your imaginary future self. Here is a diagram:
Although this example was a little contrived, it can be generalized to many
other traditional computation problems, but not all. On the other hand, there
are problems that can only be solved by imaginary computation. In that sense,
124
it is similar to uncomputation. Of course, imaginary uncomputation is useful
as well.
3
Complex Computation
To combine the unique capabilities of computation, uncomputation, and imaginary computation, we need the capability to do both at once. Fortunately, we
can do this with threads. We can set one thread to run in real time, and another to run in imaginary time, and two more to run backwards and imaginary
backwards. As you might expect, complex multithreading is an intimidating
prospect. However, unknowingly, we have already been making use of imaginary computation in video games. How else could Solid Plissken talk to Big
Snake over the codec without any time passing for the mans around him?
By studying the use of complex computation in video games like Hard Metal
Gear, we will be able to apply it to other fields as well. However, that is the
extent of my current research.
These are my references. Myself, Harry Q. Bovik for helping me prepare this
paper and teaching me how to make very good LaTeX papers, Isaac Newbton for
helping me develop a system for notationalizing the symbolology, and Hideous
Kojima for making Hard Metal Gear.
125
126
Track 1
Perplexity Theory
О
ach
SHALL I AT LEAST SET MY LANDS IN ORDER?
The Reviewers Must Be Crazy ......................................................129
C. J. Deere
О
LONDON BRIDGE IS FALLING DOWN FALLING DOWN FALLING DOWN
FAIL-Talk: Fully Automatic Incomprehensible Language Talks ..135
Alga Rhythm
О
POI S’ASCOSE NEL FOCO CHE GLI AFFINA
An Analysis and Redesign of the Evacuation Instructions in an
Academic Office Building .............................................................139
I. B. Gone, U. Cower
О
QUANDO FIAM CEU CHELIDON—O SWALLOW SWALLOW
All my friends’ researches are cooler than mine ............................143
Chris A. Maphone
О
LE PRINCE D’AQUITAINE À LA TOUR ABOLIE
How to successfully prevent the flow of information in
research presentations ....................................................................145
Kami Vaniea
О
THESE FRAGMENTS I HAVE SHORED AGAINST MY RUINS
An Alternate Construction Method for the Surreal Numbers ........153
Jim McCann
127
Track 1
Perplexity Theory
(continued)
О
ach
WHY THEN ILE FIT YOU. HIERONYMO’S MAD AGAINE.
The first paper on SAMIR – ANVESH..........................................155
SAMIR Sapra and ANVESH Komuravelli
О
DATTA. DAYADHVAM. DAMYATA.
The second paper on SAMIR – ANVESH.....................................157
О
SHANTIH SHANTIH SHANTIH
The third and the last paper on SAMIR – ANVESH .....................139
128
The Reviewers Must Be Crazy
A Plea For Sanity in Peer Review
C. J. Deere
March 13, 2010
Abstract
this disturbance in the sense of self can lead
to periods of dissociation.
While reviewing a conference paper recently, the author noted that one of the options given as an evaluation of the paper was ‘borderline’, an entirely precise
and informative assessment in a sea of unrelated measures such as ‘strong reject’ and ‘weak accept’. This
paper proposes a reviewing process that is compatible
with current psychiatric best practices.
1
Introduction
We see in Figures 1 and 2 a typical review request
form send by a PC member to a grad student who
might maybe possibly have some idea of what the
paper is about, at least enough to make a plausiblesounding case for rejecting it.
Tellingly, the option of ‘borderline paper’ comes up
all too frequently. Wikipedia, via the fourth edition
of the Diagnostic and Statistical Manual of Mental
Disorders (DSM-IV) tells us that
Borderline personality disorder (BPD) is
a personality disorder defined in DSM-IV
and described as a prolonged disturbance of
personality function in a person (generally
over the age of eighteen years, although it
is also found in adolescents), characterized
by depth and variability of moods. The disorder typically involves unusual levels of instability in mood; black and white thinking,
or splitting; chaotic and unstable interpersonal relationships, self-image, identity, and
behavior; as well as a disturbance in the individual’s sense of self. In extreme cases,
We note that indeed, ostensibly for “budgetary reasons”, very often color printing is disallowed, thus
leading to an extremely black and white expression
of the author’s intent. Most authors are over the age
of eighteen. Mood instabilities frequently result from
notification of paper rejection or acceptance. Authors constantly write ‘we’ in place of ‘I’, even in
single-author publications — this is plainly a severe
“disturbance in the sense of self”.
2
Prognosis
The general ambience of of academic writing and
publishing is plainly a psychologically toxic environment, so much so that even the designers of the review form let the fact slip out in their choice of wording. A cry for help, as it were.
We propose that reviewers should be able to
take advantage of the most up-to-date classification
schemes developed by expert psychologists in order
to assess their peers’ work more honestly. Our proposed conference review form is given in Figure 3, and
we give below representative examples of the sort of
writing that is meant to merit each assessment.
3
Examples
In this section we take
129
3.1
Manic
Interpreting linear logic as a lambda calculus yields linear function types, which are
required to use their argument exactly once.
Oh man that is so awesome. I’m gonna
to write a thousand-page novel about linear types and it’s gonna — they’re gonna
make a movie about it and I’ll be so rich
and famous and woo linear types!
3.2
3.6
Schizophrenic
Interpreting linear logic as a lambda calculus yields linear function types, which
are required to use their argument exactly
LET’S RIDE BIKES!
Interpreting linear logic as a lambda calculus yields linear function types, which
are required to use their argument exactly once. Well, that’s what the GOVERNMENT wants you to think, anyway.
They implanted contexts of linear hypotheses behind my ears. They’re controlling
my thoughts via algebras of resource possible worlds. The proof-terms tell me to do
things. Very bad things.
3.3
3.7
Antisocial
Borderline
Fuck you. Oh, I’m sorry, I didn’t mean to
say that. It’s just that I care about you so
much.
Fuck you.
3.5
Depressive
And then it’s all used up. I can never get
that resource again. What’s the point of going on, then? I loved that linear resource.
I’m never going to be able to have it, never,
ever again. Sigh.
Fuck you.
3.4
ADHD
Obsessive-Compulsive
130
Date: Thu Apr 1 11:12:09 2010
From: mr@cs.cmu.edu
To: bovik@cs.cmu.edu
Subject: BLEEP’10 #72 -- Review Request
Dear [Reviewer],
below is the review form for the paper
***
***
***
***
SUBMISSION NUMBER: 42
TITLE: On Dewubulating Widgets
AUTHORS: J. Blow
PC MEMBER: Dr. Michael Romanage
submitted to BLEEP’10.
I need your review back by email by
Wednesday Mar 31 (i.e., yesterday)
so I can integrate it with my own half-baked notions and submit to
EasyChair on time.
Thanks,
Mike
---------------------------------------------------------------- For your convenience, this form can be processed by EasyChair
--- automatically. You can fill out this form offline and then
--- upload it to EasyChair. Several review forms can be uploaded
--- simultaneously. You can modify your reviews as many times as
--- you want.
--- When filling out the review form please mind
--- the following rules:
--- (1) Lines starting with --- are comments. EasyChair will
--ignore them. Do not start lines in your review with ----as they will be ignored. You can add comments to the
--review form or remove them
--- (2) Lines starting with *** are used by EasyChair. Do not
--remove or modify these lines.
-------------------------------------------------------------*** REVIEW FORM ID: 24561::9999
*** SUBMISSION NUMBER: 42
*** TITLE: On Dewubulating Widgets
*** AUTHORS: J. Blow
*** PC MEMBER: Dr. Michael Romanage
-------------------------------------------------------------*** REVIEW:
--- Please provide a detailed review, including justification for
--- your scores. This review will be sent to the authors unless
--- the PC chairs decide not to do so. This field is required.
Figure 1: Typical Review Form, part 1
131
SUMMARY
+ Pluses:
- Minuses:
-------------------------------------------------------------*** REMARKS FOR THE PROGRAMME COMMITTEE:
--- If you wish to add any remarks for PC members, please write
--- them below. These remarks will only be used during the PC
--- meeting. They will not be sent to the authors. This field is
--- optional.
---------------------------------------------------------------- If the review was written by (or with the help from) a
--- reviewer different from the PC member in charge, add
--- information about the reviewer in the form below. Do not
--- modify the lines starting with ***
*** REVIEWER’S FIRST NAME: (write in the next line)
*** REVIEWER’S LAST NAME: (write in the next line)
*** REVIEWER’S EMAIL ADDRESS: (write in the next line)
---------------------------------------------------------------- In the evaluations below, uncomment the line with your
--- evaluation or confidence. You can also remove the
--- irrelevant lines
*** OVERALL EVALUATION:
---------------
3 (strong accept)
2 (accept)
1 (weak accept)
0 (borderline paper)
-1 (weak reject)
-2 (reject)
-3 (strong reject)
*** REVIEWER’S CONFIDENCE:
-----------
4
3
2
1
0
(expert)
(high)
(medium)
(low)
(null)
*** END
--------------------------------------------------------------
Figure 2: Typical Review Form, part 2
132
***
***
***
***
REVIEWER’S
REVIEWER’S
REVIEWER’S
REVIEWER’S
FIRST NAME: (write in the next line)
LAST NAME: (write in the next line)
EMAIL ADDRESS: (write in the next line)
CONSULTATION FEE: (enclosed)
*** OVERALL EVALUATION:
---------------
3 (manic)
2 (schizophrenic)
1 (antisocial)
0 (borderline)
-1 (obsessive-compulsive)
-2 (ADHD)
-3 (depressive)
*** REVIEWER’S CONFIDENCE:
-----------
4
3
2
1
0
(I can spell words with all the letters after my name)
(herr doktor)
(doctor)
(well my family’s pretty crazy on average)
(read something about psychology in "popular science" once)
*** PRESCRIPTION:
-----------
4
3
2
1
0
(electroshock)
(brand-name medication)
(generic medication)
(therapy)
(placebo)
*** END
--------------------------------------------------------------
Figure 3: Revised Review Form
133
ach
sigbovik
Paper 12: The Reviewers Must Be Crazy
Reviewer 1
Confidence: 4/4
The paper (which I clearly didn’t read, I just read and review things based on their bylines and titles
anyway) makes a strong case that the reviewers must be crazy. This insightful, necessary, and longoverdue conclusion is then supported by several pages of text that I didn’t read as well as a figure or
two that I might or might not have cursorily examined. The author might or might not have included a
conclusion or two as well as a call for future work at the end of the paper. If he did, I have to confess
that I really didn’t get that far. But hey, that happens sometimes, right?
As a reviewer myself, I can vouch for the author’s results and conclusions (by which I mean, his title)
since I might or might not be crazy. Therefore his work must be true since I am always right. QED.Of
course, I must follow every reviewer ever by giving the same comment: your work is excellent/good/
adequate/fair/poor/omgburnitwithfire but you need to cite more of the empirical literature. As this
review is now in the public domain, I highly encourage you to cite this and raise my citation count so
I can get tenure and stuff. You know, because I really want to be cited by a paper that suggests that all
reviewers are crazy. Deans can’t read but deans can count, right? Hey actually, that could be a paper
someday, just don’t steal that idea....or if you do, just cite me and we’ll all be good.
STORNG ACCEPT!!!11!!one one one
134
FAIL-Talk: Fully Automatic Incomprehensible Language Talks
Alga Rhythm
Abstract
FAIL-Talk
The author previously introduced Auto-TALK,
a solution for making boring, text-based talks
much more interesting [1]. A potential problem with Auto-TALK is that speakers can become too interesting and students may be requested to give many talks at seminars, taking
valuable time away from research. In this work, a
“metaphorical inverse” FAIL-Talk is introduced,
which solves the too-many-talks-not-enough-time
problem.
• Making a fun talk is hard
• Auto-TALK solved this problem
• Making too many good talks introduces new
problems
– Talks too interesting
– Asked to give too many talks
• Introducing: FAIL-Talk
– No one should ever ask again
Figure 1: Initial slide, pre-transformation
1
Introduction
No one (initially) wants to give a bad talk. In
prior work, Auto-TALK [1] was introduced a solution to the endemic problem of boring talks.
It incorporated pictures of puppies, kittens and
YouTube videos to gather and hold audience attention. An unfortunate side effect may be students being so interesting that the demand for
talk impinges on valuable research time. This
is known as the too-many-talks-not-enough-time
problem: grad students only have so much time,
and preparing talks takes time which could otherwise be spent on research.
Our proposal to remedy this travesty is FAILTalk. Like Auto-TALK, FAIL-Talk provides a
concrete set of steps (an algorithm, if you will)
to transform a talk into a new “improved” talk.
Unlike Auto-TALK, the goal of a FAIL-Talk
transformation is to discourage requests for further talks. Ever.
We will utilize simple strategies to transform
talks from basic but comprehensible to completely incomprehensible. Our toolkit is easy,
adaptable, and multi-purpose, able to be deployed cross-discipline without much adaptation.
2
FAIL-Talk Transformations
There are three main techniques for making a
basic talk horrible: talking too fast, unintelligible slides, and random incorporation of multivariable integrals.
135
2.2
Unintelligible Slides
FAIL-Talk
The less sense the slides make, the less people
will enjoy a talk and the less they’ll want to
repeat an experience. (While the author has
no proof of this statement, it appears reasonable.) So rather than black text on a white
Triangles
– Talks too interesting
Forever
background, brown text on a black background
– Asked to give too many talks
is recommended. Whenever possible, animations
• Introducing: FAIL-Talk
should be used to cover text with completely useBoom!
– No one
should ever ask again
Bang!
less block diagrams. (Note: it is important that
x y z x sin x 17 dzdydx
the block diagram be completely unrelated to the
talk).
Figure 2: After the addition of random shapes
If this is insufficient, more drastic measures
with text, an integral, and text perturbation in can be taken. Text can be perturbed with rananother language (here, mostly Greek)).
dom characters from other languages to ensure
slides make absolutely no sense.
It is important to note that stripping the talk
FAIL-Talk
of all relevant images/video and inserting bullet points randomly and with high frequency are
• M•aking a fun talk is hard
Blue is pretty!
also highly recommended.
• AUTO-Talk solved this p•••roblem
• Making a fun talk is hard
Blue is pretty!
• AUTO-Talk solved this problem
• Makßing too many good talks introduces
new problÞems
yz 2
x
y
5x y
x y z • Makßing too many good talks introd•uces
n•ew pr•oblÞems
2.3
Triangles
– Talks too int•eresting•• Forever
– Asked t•o give too many talks
• Introducing: F•A•I•L-T•a•l•k
Boom!
– No one
should ever ask aga•in
Bang!
x
yz 2
y x z y x 5 x sin x 17 dzdydx
y
x y z Figure 3: Black background, brown text, and
extra bullets for intra-word emphasis devastate
any remaining comprehensibility.
2.1
Talking Too Fast
Incorporating Multi-Variable Integrals
To complete the transformation,
some
−percentage of slides should have multivariable integrals added, such as
∞ ∞ ∞
2
y
xyz +y x z y x5x |sin(x) + 17| dzdydx
x=−∞
y=−∞
z=−∞
The harder the integrals are to solve, the better; the integrals should have no relevance whatsoever to the talk. (In the event that integrals
seem too drastic, other mathematical equations
can be inserted, with lesser effect).
The author asserts that talks become harder to
understand in proportion to the speed the pre2.4 Commutativity of transforms
senter speaks. The easiest transformation, then,
is simply for the presenter to speak as quickly as While the transforms provided may not comhe/she can.
pletely commute (example: the order which two
136
Not Inverses
Not Inverses
• Sometimes
So
ometimes it
its
ttss ha
h
hard
ard to u
undo
o things
th
• Ev
Even
ven if it seem
seems
ms like A
m
Auto-TALK and FAIL-Ta
FAIL-Talk
T lk
aree inverses,
inve
vers
ve
erses
es, d
do
don’t
on
n’’t be
be d
deceived
eecceived
POW!!!!
• They
hey
e aaren’t
ey
rreen
n’’t
• So
Sometimes
ometimes it
its
ts ha
h
hard
ard to u
undo
o things
th
• Ev
Even
ven if it seem
seems
m like A
Auto-TALK and FAIL-Talk
T lk
l
aree inverses,
inve
ver
ve
don’t be deceived
POW!!!!
P
OW!!!!
• Th
They
hey
ey aaren’t
rreen’’t
Figure 4: FAIL-Talk applied to Auto-TALK. Figure 5: Auto-TALK applied to FAIL-Talk.
Dog is now under the “Pow” shape. Again, the Original is not restored, and dog appears in front
original text slide is not restored.
of “Pow” shape.
basic powerpoint figures appear may influence
which lies on top of the other), the author argues
that the net effect of the transforms does indeed
commute. For example, block diagrams can be
added, then text can be perturbed by a foreign
language, followed by the addition of integrals.
If at this point the author were to change the
text to brown and the background to black, and
finally insert excess bullets, the progression may
resemble Figures 1, 2, and 3. Clearly, the net
effect is an incomprehensible slide, regardless of
transformation ordering.
4
Metaphorical Inverses
The author would like to assert that FAIL-Talk
and Auto-TALK are not true inverses in the
mathematical sense. For a set of slides p ∈ P ,
where P is the set of all possible PowerPoint
talks, FAIL-Talk(Auto-TALK(p)) may yield a
talk, with an embedded YouTube video which is
covered by a block diagram containing an integral. An example of FAIL-Talk(Auto-TALK(p))
is shown in Figure 4, where puppies are applied
before a large “Pow!!!” shape. Likewise, AutoTALK(FAIL-Talk(p)) may make a bad talk
somewhat more bearable (via puppies) but does
not restore comprehensibility or even the orig3 Experimental Evaluation
inal talk.1 An example of Auto-TALK(FAILTalk(p)) is shown in Figure 5, which simply inAt this point, the author has not yet field verts the ordering of the transformations shown
2
tested FAIL-Talk, as she wishes to avoid burn- in Figure 4.
ing bridges. However, mocks are included as Fig1
The author recommends keeping a backup copy if
ures 1-3, clearly demonstrating the devolution of there is a possibility of a need for the original talk.
2
a slide from one that makes sense (Figure 1) to
Note that neither Figures 4 or 5 restores the original
talk, which was a pure text slide.
one which is horrifying (Figure 3).
137
5
Conclusion
This work presents FAIL-Talk, a fully automated
scheme to make incomprehensible talks. While
the automation is not yet complete, the author
has demonstrated the awe-inspiring power of this
scheme to completely destroy the substance of a
talk while adding only nonsense via a few simple powerpoint mocks. While this work can be
thought of as a solution to the problems AutoTALK created, it may itself yield side-effects:
presenters should be prepared to never be asked
to give a talk again, even if they eventually want
to.
The End
References
[1] Alga Rhythm. AutoTALK: An Automatic
Presentation Graphical Toolkit. In SIGBOVIK, 2009.
138
An Analysis and Redesign of the Evacuation Instructions
in an Academic Office Building
I. B. Gone and U. Cower
Abstract
An evacuation process is provided for an academic office building. The efficacy of this
process is evaluated, and the maturity of the evacuation model is assessed to be EMM Level 1.
A scalable improved process description is proposed.
Introduction
Safety is recognized as an essential
human need and a prerequisite for
productive activity. Safety needs are
placed just above physiological needs
and below belonging, esteem, and selfactualization by Maslow’s hierarchy.
A major university, in its infinite
wisdom, has provided a “quick
reference” guide to the evacuation
process for its academic buildings. The
evacuation process, as promulgated to
building occupants, is exhibited in
Figure 1. The manifest complexity of
this description has led to postings with
supplemental instructions, such as the
example of Figure 2.
Figure 2: Supplement to Evacuation Process
Figure 1: Official Evacuation Process
139
Despite the distribution of this descriptive quick reference poster, lack of familiarity with the
process on the part of building occupants is suggested by considerable anecdotal evidence. After
all, the number of people who will actually memorize this stuff is vanishingly small, and who’s
going to read the fine print with flames licking at their ankles?. Unfortunate consequences could
ensue in the event of an emergency.
A user study was conducted to determine the efficacy of the evacuation process. The results
were assessed under the Evacuation Maturity Model, with the result that the process is at EMM
Level 1. A revised, scalable description of the evacuation process is proposed.
Efficacy of the Evacuation Process
A number of building occupants were recruited dragooned into this study. They were
presented with the current evacuation process, their interpretations were observed, the data was
de-identified, and they were then placed in simulated emergency situations (Figures 3, 4, 5).
As can be seen from the illustrations, confusion and consternation were prevalent. The
principle of intra-ocular trauma is satisfied by the data, so no further statistical significance
testing is required.
Figure 3: Simulated bomb emergency
140
Figure 4: Simulated fire emergency
Figure 5: Simulated weather/flood emergency
141
Evacuation Maturity Model Assessment
The Evacuation Maturity Model (EMM) was developed as a tool for objectively assessing
evacuation plans of businesses and institutions. A 5-Level process maturity continuum is
recognized, where the uppermost (5th) level is a notional ideal state where processes would be
systematically managed by a combination of process optimization and continuous process
improvement. The five levels are defined as:
1.
2.
3.
4.
5.
Initial (chaotic, ad hoc, individual heroics) - the starting point for use of a new process.
Repeatable - documented procedures and checklists are used to guide the process
Defined – confirmation as a standard business process is accomplished
Quantitatively managed - quantitative goals and objectives are incorporated
Optimized - deliberate process optimizations/improvements are included
Lack of understanding of the process on the part of the occupants was revealed by the efficacy
study. Study of the process would substantially lengthen the critical path to emergency action. In
the event of an emergency, heroics would be required. Explicit objectives are lacking. A maturity
level of EMM1 is assigned by the maturity assessment.
Improved Process Description
An improved process description design is suggested by the data. Details of the process, even
if required for precision, are elided. Comprehension and interpretation tasks are removed from
the critical path in favor of action. An approximate process actually executed is deemed
preferable to a precise process ignored.
The improved process description is presented in Figure 6. Dimensions for the document are
not provided, as the description is intended to result in evacuation process improvement at all
scales from wallet card to billboard.
if
Fire
Gun
Storm
Flood
No pulse
Blood
then
Run
Hide
Stay
Climb
Zap
911
Figure 6: Scalable simple evacuation process description
142
All my friends’ researches are cooler than mine
Chris A. Maphone
March 21, 2010
Abstract
We have done no research and have nothing to say about it. Instead this paper presents solutions
a multitude of exposition problems: we take it upon ourselves to succinctly and accurately describe
the research of our friends and colleagues, to the best of our understanding. We think that previous
work (i.e. their own explanations of their work) vastly understates how TOTALLY AWESOME
what they are doing is and misses key points about real-world applications of their work for e.g.
blowing stuff up.
1
Introduction
tion of the typographic characters in the formula. Then his labmates turn these animations
This is a survey paper on some of the recent into robots. The robots are made with certain
exciting research endeavors of many awesome “gates” (like and and not) described by the equaCarnegie Mellon students and alumni!!!!!!!!
tions. One of them has VERY SHARP BLADES
and has to be kept in a box. Someday, these
robots will get very tiny and invade our bodies
2 Research
for medical science.
Dan Licata works on something called “dependent types”. They are so powerful they can do
ANYTHING. He is working on building the ultimate type theoretic weapon which uses focussing
to create a concentrated beam of power in the
form of two “arrows”, “Mr. Annihilate” and “Dr.
Destroyer” [1]. This weapon is called a TASER
(type amplification by stimulated emission of radiation).
Ross Hatton works on kinetic typography,
which is like letters moving around and changing size and stuff. Specifically he applies the
kinetic approach to calculus formulas and, using Stokes’s theorem, can describe the very mo-
143
William Lovas refines types finer and finer until they are sharp enough to stab a bitch [2], and
then some more until they are a soft powder.
This means you can describe your data so precisely that it requires quantum physics, so like if
you have a natural number it’s actually a quantum superposition of even and odd and you never
know for sure until you observe it with type reconstructivism.
Michael P. Ashley-Rollman (MPA) makes
blocks that light up. I think they are children’s toys, but I like playing with them too even
though I am a mature grown-up person, pretty
weird I know. Anyway you click them together
and they change colors, and sometimes because
of racial conditions the colors spazz out and start
flickering a lot. Michael is working on solving the
problem of racism.
Luke Zarko is probably a superhero otherwise why would he have that name? He uses
his superpowers to make an operating system
(ZARKOS) that already knows at compile time
what your code is going to do, so you never
have to run anything and it’s very safe. Unsafe
programs like GUN BASH never even get executed in the first place because you can tell from
the name that it’s very fangorious. This operating system is not only provably correct but also
VERY FAST.
Rob Simmons makes logic programs go backwards. It’s actually kind of confusing because
usually “backwards” means “forwards” so sometimes they say “bottom-up” which I think is related to his prior SIGBOVIK publications on
drunken logic [4]. Anyway every time he makes
a program go backwards he gets a new elf:
Tom 7 used to make logics about spaceships
but now he uses machines to learn things. His
thesis is titled “Model Types for Mobile Code”
[3] and it has to do with model logic (as in model
spaceships, because real ones are too big to fit in
a computer, even a distributed one).
144
3
Conclusion
I JUST HAD TWO ESPRESSOS!!!!!!!!!!!!! The
end.
References
[1] Daniel Licata. The strange case of dr. admissibility and mr. derive. Speaking Skills Talk
at CMU Student Seminar Series, 2008.
[2] William Lovas and Frank Pfenning. A bidirectional refinement type system for LF.
Electronic Notes in Theoretical Computer
Science, 196:113–128, January 2008.
[3] Tom Murphy, VII. Model Types for Mobile
Code. PhD thesis, Carnegie Mellon, January
2008. Available as technical report CMU-CS08-126.
[4] Rob Simmons. A non-judgmental reconstruction of drunken logic. In Proc. SIGBOVIK
2007. ACH, April 2007.
How to successfully prevent the flow of information in research
presentations
Kami Vaniea
March 21, 2010
Abstract
The main body of research presentation literature is dedicated to the creation of great research
talks in which the presenter successfully conveys
their great wisdom to an audience of receptive
minds. The great art of preventing your audience from figuring out what you actually did is
not nearly as well documented or explored. In
this work I will explore the different presentation
mechanisms that can be used to simultaneously
give an excellent talk, thoroughly confuse the audience and leave them with a feeling that you
really understand what you are talking about.
1
Introduction
Creating a hard-to-comprehend talk is amazingly similar to creating an enlightening talk.
You need to do things like consider your audience, plan out what you are going to say and
meticulously create your slides to guide your audience through your material. The point where
the two methodologies diverge is in the goals.
Unlike a traditional information imparting presentation an information impeding presentation
has the dual goals of convincing your audience
you know what you are talking about while makings sure they don’t understand anything you
say.
Some critics may say that creating information
impeding talks is a debasement of science and a
complete waste of your audience’s time. However, they are overlooking the needs of thousands
of students who’s results were neither interesting nor useful placing them in the not interesting and not useful results quadrant, aka Bovik
quadrant, of the quest for fundamental understanding and considerations grid put forth by
Donald Stokes [2]. Unlike the praised work in
Pasteur’s quadrant the work in Bovik quadrant
is woefully ignored and not given the recognition
it deserves.
The purpose of this work is to assist students
everywhere in accomplishing two goals with their
talks 1) convince the audience that you did
something worthwhile 2) make sure the audience
never figures out what that important thing was.
These are ambitious goals but never fear they
are completely accomplishable if you pay close
attention to this guide.
In order to impede information flow during
your talk you need to accomplish the following
sub tasks:
• Encourage interior audience discussion
• Belittle the audience
• Mislead the importance and relevance of
your data
• Use as much time as possible explaining
unimportant information
In this work I will discuss each of the major components of a talk and look at different ways to
accomplish these tasks in each component. Curious how your going to accomplish all this in a
single talk? Well, read on.
145
2
Motivation
Cloud computing! Cancer! Google! Health care!
Kittens! Global warming! Remember, whatever
your research is it can be related to the hot news
item of the week. All motivation really is tying
your research to your favorite well funded buzzword. Remember that good motivating examples accomplish three things 1) convince audience your work is important 2) take up as much
time as possible and 3) make it unclear just how
many of the world’s problems you are actually
going to solve in this talk.
When properly obfuscating a talk its important to pick your motivating example carefully.
You want an example that lots of people care
about so that they will leave your talk feeling
like you solved something important even if they
didn’t understand a word of what you did. But
you also need a motivation that not too many
people understand and is hard to ask detailed
questions or point out inconvenient facts like
your motivation and your work have nothing in
common.
Cloud computing is a wonderful example of
a good motivation topic. Everyone knows that
cloud computing important, complicated and either doomed to failure or the next bit thing.
Cloud computing needs machine learning to ensure optimal usage. It needs security to protect
data from the unnamed dangers lurking in the
clouds. It needs theory if it is ever going to work.
It needs PL to specify how different elements can
interoperate safely. Whatever your work is about
play it up in relation to your motivating topic.
Without your vital, unspecified, work your motivating topic would be crippled and the world
would descend into chaos.
When introducing a motivating example, like
cloud computing, be sure to spend plenty of time
on it. Remember the more time you spend on
your motivating example the less you have to
talk about your actual research or even worse,
your results. Besides motivating examples are
far more fun to talk about and listen to. So
spending lots of time on it will actually be en-
joyable for both you and your audience.
One of the purposes of a motivational topic
in an information imparting presentation is to
introduce the problem you are going to be solving in the remainder of the talk. In an information impeding presentation this is an excellent
opportunity to mislead your audience by implying that your research solves many more problems than it actually does. For example strongly
type checking the Internet would make it more
secure. Therefor nearly all security problems associated with cloud computing can be solved by
applying a novel type checking method.
3
Problem statement
The problem statement is arguably the single
most important part of your talk. The problem statement lets your audience know what you
will be talking about. This is bad. You do NOT
under any circumstances what the audience to
figure out what you are talking about.
If possible, you should really skip this part of
your talk. The best solution is to leverage your
motivating example and just assume that it doubles as a problem statement. Thereby not skipping the problem statement while still not clearly
explaining it.
4
Outlines and organization
Outlines are generally a bad idea if you want to
make sure your audience gets the wrong lesson
from your talk. But this doesn’t mean you can’t
have an outline slide. Remember the goal here is
to give what looks like a good talk while ensuring
the audience doesn’t understand what you say.
A good outline will let the audience know what
is coming, what you are going to tell them about
and how you are going to solve your problem all
in a neat package. Creating an outline with no
meaning that meets these requirements is actually trivial just copy the following onto an outline
slide. In your talk just read the slide and provide
no extra information.
146
• Introduction
• Motivation
• Background
• Methodology
• Results
• Conclusion
This simple outline gets you points with the
audience for having an outline while giving them
no additional data about what is coming.
5
Graphs and diagrams
Figure 2: Example of a graph showing the clear
difference between your good result and the opGraphs follow a similar rule to statistics. Re- posing technologies poor result. Remember clear
member there are lies, dam lies and statistics. results speak for themselves and if they don’t you
Similarly graphs give you the ability to give your just need to muddle with the graph until they do.
audience the wrong impression while not lying.
To begin with there are the tried and true
If simple graph alterations fail to adequately
graph and figure obfuscation techniques. Leave
describe
the importance of your results you
out labels. Leave out the scales. Use three dimensions. Use as many of Excel’s colors, tex- may have to resort to more complicated tricks.
tures, and eye candy as possible thereby impress- Thankfully many books have been written on
ing your audience with the visual complexity. A how to build good graphs which also convegood example of this sort of graph can be seen niently show lots of examples of bad graphs.
in Figure 1. This is actually a graph of all the Tufe’s famous book on The Visual Display of
book sales in the US from 1940 to 1989 but that Quantitative Information [3] is an excellent reis impressively hard to tell from the graph. Use a source.
Chinese coloration scheme where red is good and
white is bad which completely confuses Ameri- 6 Audience Questions
can audiences. Put in as much data as possible
even if it could be left out or isn’t relevant to Annoyingly, audience members are typically alyour point.
lowed to ask questions during and after your talk.
Math is already a horribly confusing topic why Life would be much easier if this practice could
would any sane person put more of it into a talk be avoided but unfortunately it is ”traditional”
on purpose? Be wary of adding math and es- and cannot be prevented.
pecially wary of adding statistics which can be
Much like an information imparting talk it is
questioned. It is far better to add a graph sim- important to anticipate audience questions in adilar to Figure 2 which shows the clear superior- vanced and have prepared answers. Knowing
ity of your technology. If the difference is clear your answers in advance allows you to answer
enough to your audience you can safely skip over questions in a way that doesn’t answer the quesany issues of statistical significance or error bar tion while making your words drip with confireporting.
dence and certainty. Much like a lion tamer you
147
Figure 1: Example of a visually confusing graph created using simple MS Excel default tools. Note
the lack of any explanation as to why this graph is here or what it is about.
148
should show no fear before your audience and
subdue them through your aggressive eye contact and confidence.
If an audience member asks a question which
reveals a flaw in your work or in some way
doesn’t show you in the best light, answer a different question. Beauty is in the eye of the beholder and question interpretation is in the ear
of the presenter. If you need some examples of
how to do this effectively just watch any presidential debate. The art of question avoidance
and deflection is what politicians are very good
at. If you need to cite facts in your answer pick
little known works and refer to the author by
first name. (For example: Joe)
Get upset when others challenge you. An attack against your research results is really a personal insult to you. After all you poured your
heart and soul into this research for the last n
years who are they to condemn it after less than
an hour? Also, a heated discussion between you
and the audience member will distract the rest
of the audience from your talk and reduce the
time available for other people to ask questions.
If an audience member asks you a question
about a result which isn’t yours you should defend the other author and answer as if it was
your research. This shows deference to the other
author’s hard work and derails the talk for several minutes until the audience discovers this is
cited work not yours.
7
Belittling an audience
for your audience to appreciate just how much
work you had to do and how mentally taxing
and frustrating it was. How do you do this? Simple, show them the details. Don’t hold anything
back. If there is something confusing in your
domain don’t be afraid to say it. If you have
tons of data points show them all. And most
importantly do not simplify or explain yourself.
You are the Alan Turing in the room and they
are supposed to keep up. And if they can’t keep
up that is an indication of their ineptitude not
yours.
Frequently use terms like ”obviously” and
”clearly” to underline how little your audience
actually knows. Obviously your research cures
all the worlds problems. Your results clearly
show this and all the equations and math backing
it up are intrinsically obvious. Don’t bash this
into the audience. A truly superior researcher
just knows that they and their research is perfect. If you approach your talk with this in mind
and a good bit of confidence you will be impressed at how fast your audience falls into line.
One simple trick is to use ambiguous domain
specific words. Everyone knows that people who
use large uncommon words are smarter. If your
audience has to spend time figuring out the
words you are using then they will miss some of
the results you didn’t want them to understand.
A good example of a seemingly innocent word is
”feature”. Which when used in a Machine Learning context has a specific general meaning as well
as a specific meaning in your research which you
can conveniently forget to provide thereby making the audience feel inadequate for not understanding this clearly obvious term. Additionally,
audience members from outside the domain will
assume they know what the term means and will
stay confused for many slides before giving up
hope of understanding the clearly complex research you are engaged in.
You are smart. Well that is obvious since you
are reading this paper. Its important to let your
audience know how smart you are. Since you are
reading this paper your results likely fall in the
Bovik quadrant [2] and are unlikely to impress
upon your audience how smart you are. Therefore its vital to make this fact clear in other ways.
If you have equations with 20+ variables that
To make your audience understand how smart
you are it is vital to make them understand just are impossible to comprehend without serious
how much they don’t know. After all getting study, don’t be afraid to show them. Math equathe results you have was hard. Its important tions are a truly beautiful construction. They
149
are arguably competently self explanatory and
therefor require little to no explanation. They
also require a significant amount of attention and
brain power to comprehend and map onto your
talk topic. A sufficiently high number of complex equations and/or proof statements will often cause a portion of your audience to give up
and just assume that you know what you are
talking about.
Graphs are very useful in belittling an audience. However, they are such an important topic
that they appear all on their own in Section 5.
8
Practice Talks
Giving practice talks to ensure minimal information conveyance is typically frowned on. Really if
you have time to give a practice talk you should
have been using that time to get better results
and therefor not need this guide. However, if you
are determined to practice your non-information
giving skills there are a few things you can do to
make your practice talk more successful.
Give your practice talk audience slide handouts to make notes on. Having something to
write on makes the audience members feel more
important. Don’t worry, if you have picked your
audience well most of them won’t actually write
anything. The handouts are also useful for you
to go back through and read the notes. Comments like ”where is this going?” or ”??!!??”
or better yet ”impressive!” indicate that your
talk is going well. While comments like ”why are
you using the XYZ methodology instead of the
ZYX methodology?” indicate a dangerous level
of methodology understanding in your audience.
Only invite people who already know your research. This way they don’t actually have to
listen to your talk but can give good advice at
the end and explain to the other audience members what you actually did in far more glowing terms than you could because unlike you
they can exadurate in ignorance. They are also
less likely to ask hard questions or nit pick your
slides.
Invite the people who came to your practice
talk to your actual talk. That way if anyone
complains about the talk they will be on hand
to explain just how bad it used to be and how
impressed they are with your improved presentation.
9
Concluding the talk
Writing a talk conclusion is actually extremely
easy and nearly identical to an information imparting talk. Remember that outline slide we
created in section 4? Take the same set of bullet
points, toss out the Introduction and Conclusion
and make each point into a partial sentence. The
resulting set of bullet points should read something like this: I told you about this big important problem (Motivation) that no one else
has adequately solved (Background), I did something brilliant (Methodology) and completely
solved the problem (Results).
Remember to NEVER, EVER introduce new
material in the conclusion. Besides the obvious
problem of finding more material to add and the
risk of an audience member calling you on something, you aren’t even supposed to be putting
material here. Why would you make life harder
for yourself than it needs to be? The audience
doesn’t want new info, you want to tell the audience as little as possible, everything is good,
don’t rock the boat.
This is also a good place to remind you that
one of the goals of a good information impeding talk is that the audience is happy and feels
like you are competent and tried to give a good
talk. The conclusion is your last chance to make
it clear to the audience that you really did something useful, even if they didn’t understand it,
and that they haven’t wasted the last hour of
their lives. The best part is that you can do
this without ever stating anything specific. In
fact the exact words ”I showed that our solution is vastly superior to existing solutions” can
be safely said when reading the ”results” bullet
of the conclusion slide in the majority of talks.
150
What you proved or how well it worked need
never even be mentioned.
10
Simple Tips and Tricks
disagree with you. But be wary of too much
eye contact though as you can inadvertently
engage your audience and keep them from
their mid afternoon nap. Be careful who you
look at in the audience and avoid anyone
who looks distracted or who might zone out
or doze off if left alone.
The lessons above will give you a good start to
creating an optimal information impeding talk
but there are a few extra embellishments you can Talking fast This is an excellent plan if this
add that will make your talk run all that much
is how you plausibly normally talk. If you
rougher.
tend to be high strung and talk somewhere
near the speed of light there is no reason to
Page numbers Leave them out! Page numslow down for your talk. The faster you talk
bers allow your audience to make effective
the less likely your audience will comprenotes and to refer back to specific points in
hend what you are talking about. If you talk
your talk during their questions. Without
calmly normally you can still safely speed up
page numbers the audience is forced to write
your speaking by a small margin and blame
more vague notes or copy the titles from
it on being nervous.
your sides which takes more time. Also referencing the slide again during the questions Um, like, uh Talking fast is hard, talking to
slowly makes you look, well, slow, so what
will take time which decreases the number
are you to do? The answer is to add filler
of questions you will get.
words. Similar to the filler found in meat
Distracting computer messages Leave
products, word filler has no value and has
them on, thereby ensuring those audience
an unproven negative comprehension effect
members who have managed to follow
on the audience. If you are stuck, need time
your work are temporarily distracted and
to think, forget what to say, or just need to
are unable to pick up the train of your
waste more time on a slide feel free to add in
talk again.
Its best if the distracting
liberal amounts of any of the ”um,” ”like” or
messages are innocent looking and therefor
”uh” filler words. These words add nothing
not something you may have anticipated.
to your presentation but they do make it
Email notifications are good by they only
a bit harder for the audience to determine
pop up in the corner and vanish on their
what you are saying and they ensure that
own. System messages such as failure to
you are saying something even if you don’t
backup or out of battery power are much
have anything to say.
better as they cause Power Point to exit out
of the slide presentation all together buying Cluttered slides Do this. While you don’t
want to assist your audience in understandyou several minutes of technological fixing
ing exactly what you did you also don’t want
during which you don’t have to present.
to be seen as leaving out important inforAlso, the interruption is sufficient to cause
mation. The simple solution is to hide the
your audience members to loose their train
needle of knowledge in a haystack of words.
of thought.
The more words on the slide the less likely
your audience is to actually read anything.
Eye contact This is a tricky one. Effective eye
So bury your statistical significance test recontact is important during your talk. You
sults in the middle of a paragraph of text
need to use aggressive eye contact to stare
and feel confident that you both provided
down the opposition and challenge them to
151
them to your audience and that your audi- providing recommendations on how to create an
information impeding talk. Using the skill set
ence will never find them.
gleamed from this work the reader should now
Talking to the screen instead of the audience
be able to 1) convince an audience that you did
Don’t do this. While the projection screen something worthwhile 2) make sure an audience
or even your computer screen may seem less never figures out what that important thing was.
intimidating than your audience you really
shouldn’t talk to them. It can make you
look a bit crazy and imply that you are not 12 Bibliography
the top notch researcher you are. The goal
of your talk is to impress your audience References
with your supposed research skills, not
convince them that you are crazy and think [1] F. Herbert. Dune. 1965.
that projector screens can hear you and
[2] D. E. Stokes. Pasteur’s Quadrant: Basic Scitalk back. Take a deep breath and repeat
ence and Technological Innovation. Brookthe following litany against fear to yourself,
ings Institution Press, 1997.
turn around, and speak to the audience.
[3] E. R. Tufte. The Visual Display of QuanI must not fear.
titative Information, 2nd edition. Graphics
Fear is the mind-killer.
Press, 2001.
Fear is the little-death that brings
total obliteration.
I will face my fear.
I will permit it to pass over me
and through me.
And when it has gone past I will
turn the inner eye to see its path.
Where the fear has gone there will
be nothing.
Only I will remain [1].
11
Conclusion
Every year thousands of students woefully discover that their research results fall into the unprestegious Bovik quadrant. Unlike Pasteur’s
quadrant or Bohr’s quadrant, Bovik quadrant
research creators would rather no one ever discovered that their research results are both not
interesting and not useful.
Bovik quadrant members have been woefully
neglected by research talk design experts who
focus nearly all their energies on how to create information imparting talks for those lucky
researchers with good results. In this work I
have attempted to address this inadequacy by
152
An Alternate Construction Method for the Surreal Numbers
Jim McCann∗
(Image created by wikimedia commons user Acdx.)
∗ e-mail:
jmccann@cs.cmu.edu
153
154
The first paper on
SAMIR – ANVESH
{ssapra,anvesh}@cs.cmu.edu
Abstract
We introduce A Novel Verification Environment for Software and Hardware (ANVESH), which employs Stateflow Abstraction Modeling for Information Retrieval (SAMIR) to retrieve information from
the Internet, make state machines, and model check them. We employ two novel approaches to achieve
this. In this way, astonishingly, we verified everything reachable via the Internet. Results are presented
on a 42 Hz Linux box (we were simulating a machine using Simics, and our license didnt let us go any
faster). We thus, also, propose the end of the Verification era. Thank you very much, Model Checking,
for leading us this far! You will be remembered for ever!
1
Introduction
We present the last and the greatest paper on Model Checking [5]. We describe our novel mega-tool,
Stateflow Abstraction Modeling for Information Retrieval (SAMIR), which uses novel approaches of lurking
and larceny (as part of A Novel Verification Environment for Software and Hardware (ANVESH)) to acquire
and verify everything reachable via Internet thus satisfying the ultimate and long-cherished goals of Model
Checking. This is the long awaited victory of Model Checking. Hereafter, we refer to our tool as the novel
mega-tool.
A critical feature is the ability to embed novel free-trace webcams into the existing Cloud1 -computing
infrastructure using “Jack and the iBeanstalk” technology.
We guess, there is nothing much left to introduce. Without asking you to hold your breath for more
time, let’s dive into this great work.
2
Related Work
People have earlier worked for quite long on what they thought were the state explosion [1], cloud computing [2] and online model checking [3] problems. But these are completely artificial as we show in this
ground-breaking work. To be very frank, no state exploded in state explosion, cloud computing was not
on real clouds and online model checking was not online but done when Internet was not connected at all!
The goal of this paper is to ignore all this prior “work” and to explore the real meanings of these terms, to
seek out new devices on the Internet and model check them, to boldly go where no model checker has gone
before!
What State Explosion actually is . . .
3
One major difference we have in real world model checking with the prior work is that state explosion in
the prior work was completely artificial, bah! We tackle much more realistic and terrifying (and terrific
1
someone may need a reminding that real Clouds still exist
155
Figure 1: Poor battery. We feel so sad for it . Hey battery! We know, you are not the culprit! Anyway, the important thing
here is that the firmware died a brave death. Long live firmware!
) scenarios. For example, we once had an intelligence information2 in 2005 that our novel mega-tool
sneaked into a switched on Nodia 2600 firmware and started model checking! The information also said it
promptly found a serious bug in the firmware and was ready to fax a counter example to our headquarters
at the Mariana Trench using the inherent satellite communication component of the firmware of the device
(which was, by the way, captured and model checked by our novel mega-tool before to be on the safe side).
But the firmware was smart enough to realise it was about to fall into disrepute and rather than face the
inevitable shame, decided to sacrifice itself by sending the battery the following code fragment
begin
panic
xplodplease,pls,pls,...
end
This, apparently, resulted in the situation shown in Figure 13 . (If you are wondering how we got this
pic, we were lucky enough to have our novel mega-tool already sneaked into (and of course, verifying) the
Closed-Circuit TeleVision which was present in that room. FYI, the CCTV was model checked successfully!)
As we see from this specific example, the world is not ready to face the shame of getting its bugs
exposed. State explosion, is henceforth, redefined as the state of the world where an actual explosion takes
place when our novel mega-tool is ready to report the bugs!
To give this paper its dignity of definition, we pause our discussion here and continue on the next paper.
2
3
Due to the sincere work by SANIA (SAMIR-ANVESH Information Agency)
Figure courtesy: http://www.techgadgets.in/
156
The second paper on
SAMIR – ANVESH
Abstract
Please refer to the first paper on SAMIR-ANVESH.
But, some reader may say that, a part of the world may not care about its bugs. To make the world look
uniform, our novel mega-tool exploits the bugs to sneakily elevate its privilege within the buggy devices in
this shameless part of the world and induces a state explosion as implied by the following lemma.
Lemma 1 Given a device in the shameless part of the world, our novel mega-tool can sneakily elevate its
privilege which enables state explosion.
Proof. Let d be a device in the shameless part of the world. Because d is shameless, it got to have
some bug b for which it is not shameful. Given the novelty of our novel mega-tool, it will promptly find b.
We stress, d is shameless. This implies d does not feel shame if its privileges can be stolen. And we stress,
our novel mega-tool is novel! The rest of the proof is obvious.
4
Model Checking using Clouds
In this section, we describe our infrastructure using Clouds (mind you, the approach is novel enough to use
the real Clouds). We also present our two major novelties, namely, the Lurk-based sneaking (Section 4.1)
and the Theft-based sneaking (Section 4.2) which exploit this infrastructure to model check everything
reachable via Internet. We first prove a useful lemma.
Lemma 2 ∀t : time(survive(Internet, t) ⊃ ∃c : Cloud
monkeying around(c, t))
Proof. Proof by complete structural induction on the graph of the world (technically we consider its
MST rooted at Pittsburgh):
Base case: Pittsburgh
As long as there is Pittsburgh, there will be Clouds. So, we can easily find a cloud which monkeys
around the world. And Pittsburgh isn’t going anywhere, anyway! So, holds for all t.
Inductive step: Suppose the hypothesis is true for all graphs smaller than given graph T . Want to show:
claim holds for T .
Note that the base case (Pittsburgh) isn’t going anywhere. So we have plenty of clouds for everyone
in T . Rest of the proof same as the base case.
157
Now that, we do not to worry about the existence of Clouds, we explain our novel way of use of them.
Inspired by the free-food webcam, generously donated by Luis von Ahn and operated by Dec/5 Inc., we
introduce the novel concept of free-trace webcam. A free-trace webcam is installed using the fabled “Jack
and the iBeanstalk” technology. Initially, a robo-Jack installs a free-trace webcam on top of the iBeanstalk
when it just sprouts. As the iBeanstalk grows, the webcam inches more towards the Clouds (see Figure 1)1
and when the iBeanstalk ultimately reaches a Cloud, the webcam glues itself to the Cloud. Whenever the
Cloud makes a move, it gets detached away from the iBeanstalk. These special webcams are manufactured
to be solar-powered to avoid exorbitant losses to Do-cane Light Company and equipped with wireless World
Area Network cards to avoid exorbitant losses to (cast)Com. This webcam keeps recording the trace as
the iBeanstalk grows.
Figure 1: Our webcam happily rising to the Cloud after a robo-Jack installed installed it on the iBeanstalk.
4.1
Major Novelty 1 - Lurk-based sneaking
One of the key components of our novel mega-tool is property inference via a Lurk2 -based sneaking. The idea
behind this approach is simple and yet elegant (not to mention, novel and to the best of our knowledge,
a first attempt to think outside previous boxes; and our knowledge is the best, to the best of the our
knowledge, lol ). To make the presentation (too!) simple and yet elegant, we quote below a message sent
by our tool while using this novel approach.
Yo! Waz up? Well, I am just a-lurkin’. These brainless guys (traces, though) are
runnin’ and runnin’ far too lon’. Looks like I don’t have anythin’ to report yet! Sorry
buddy! I should wait till they are done and then, you know, I shud prolly get busy
inferring the damn properties and only then, can I get into my actual business. So..I’ll
keep lurkin’. Cya.
In a formal lingo, what this message is trying to say is, the tool lurks around a host, observing the
execution traces (we call this novel mega-technique, observing the observables) generated while the host
keeps running. Once the host is done, the tool uses learning techniques to infer properties that the host
has to satisfy. And the business our tool was talking about is to lurk around more and wait for observables
which refute these properties. We pause here to continue the discussion in the next paper.
1
2
Figure courtesy: http://www.teachershelp-teachingresources.co.uk/
Oh! which, by the way, means to lie in wait, lie in ambush, behave in a sneaky and secretive manner
158
The third and the last paper on
SAMIR – ANVESH
Abstract
Please refer to the first paper on SAMIR-ANVESH.
How does the tool know about the host? We use our novel free-trace webcams to eagle-eye for free
execution traces generated by hosts on the Internet. As an immediate example of lurk-based sneaking, we
recall the technology used for transporting the webcams on top of the Clouds. We noted before that, a
webcam records the trace as the iBeanstalk grows. Now, once it gets glued to a Cloud, it hands over the
information of this trace to our lurk-hungry tool which immediately detects that the iBeanstalk is going to
generate a memory-out-of-bounds exception. An iBeanstalk is humble by construction. It gets ashamed
of itself and creates a state explosion, killing itself. (By the way, this also removes a huge obstacle for the
webcam’s view!)
4.2
Major Novelty 2 - Theft-based sneaking
As a useful side-effect of our free-trace webcams, our tool gets the novel information about the sneakable
devices reachable via Internet. Then, we use the novel technique of theft-based sneaking [4] and novelly
apply a novel application of it by sneaking into the device and stealing some CPU cores for model checking.
It also novelly elevates its privileges to be able to muck around with the firmware to build a Kripke structure
(this is as good as it gets in Model Checking) and model check for various safety and liveness properties.
To show the applicability of this novel theft-based sneaking, we present the best example model checking
could have ever achieved. This is a report from one of our faithful (and grateful) sufferer of our novel megatool, in her own sweet words (and also melodious, but unfortunately, we are unable to provide the voice
memo),
One fine day, this awesome thing happened. I was checking my email on my iPhone and I suddenly get
this message flashed on top of the screen - “Congratulations! Your smartphone is now completely verified!
Don’t worry about any further bugs either in the hardware or in the firmware or in any of the apps you
might be using or even any of the updates you might possibly have in future! Your device is completely bug
free. 100%. Enjoy using your iPhone. SAMIR and ANVESH!” Simply brilliant! I don’t need to worry
about bugs in my device at all! I love Apple!
She also sent us a proof-by-image (see Figure 3)1 .
Unfortunately, the poor customer thinks this is a miracle of Apple, but we do not mind.
5
Side result
Theorem 1 Eventually, the world will by completely bug-free. Indeed, we are currently in that time.
Proof. Drumroll... Trivially true!
1
Figure courtesy: http://mactoids.com/
159
Figure 3: The happy message shown to the user by our mega-tool! This manner, we made many people happy. Yo SAMIR
and ANVESH!
6
Future Work
Wth? No future work left! Go and sleep!
7
Conclusion
We successfully deployed our novel mega-tool to model check everything reachable via Internet after successfully bootstrapping on our dear machine. See abstract for details of the machine. Hats off to Model
Checking! You did it!
Acknowledgement. We wish to acknowledge the organizers of SIGBOVIK for giving us this wonderful
opportunity to check models on the Internet. We also want to acknowledge Jim Skees for giving us
Facilities, and the SCS Help Desk for the dearest machines, Guy Blelloch for the fancy building where our
ground-breaking research was conducted, and Prof Todd Mowry for sneaking ECEs Simics [6] license into
the building (inspired by our novel application of theft-based sneaking). Refer to the rest of the doc for
possible inlined acknowledgements.
References
[1] Peter Bovik and Harry Q. Bovik: State explosion and beyond, In the proceedings of the 1st International Conference on State Explosion, Bovik Publishers, Pinatubo, 15 June, 1991.
[2] Mr. Fredrickson: My experiments with cloud computing, In the proceedings of the −29th Atmospheric
Conference on Clouds, Disney Publishers, South America, 1980.
[3] Ms. Confusee: Checking Models as they are being generated, In the proceedings of the 100th Offline
Conference on Online Model Checking, Offline Publishers, A Confused Place, 2003.
[4] Jim McCann and Ronit Slyper: A theft-based approach to 3D object acquisition, In the proceedings of
the 6th Binarennial Conference in honor of Harry Q. Bovik, 2007.
[5] http://en.wikipedia.bovik.org/wiki/Model checking/
[6] http://www.simics.net/
160

2010 PDF - SigBovik

Transcription

Similar documents

Transformers Cybertronian Bumblebee 34084 Instructions

Emergence of spontaneous proto

Hofbrauhaus - Pittsburgh Ski Club

Vladimir Luxuria Syracuse University in Florence

Faculty Finds - Cypress College Library

MS Robot for Automated Powder Coating

Fun STEM Lessons for Preschoolers

EECS490: Digital Image Processing

M-900iB_700 - FANUC America